Erratas de las fuentes publicadas

Implementar una norma en sala limpia significa volver a deducir cada fórmula, constante y ejemplo resuelto a partir del documento fuente y no del código de otra persona. Hecho sobre cientos de documentos, ese proceso encuentra defectos en las propias fuentes: un ejemplo resuelto que contradice su articulado, una constante a la que la composición tipográfica le comió un dígito, una referencia cruzada que apunta a la ecuación equivocada.

Esta página es el registro de esos hallazgos. Cada entrada nombra la edición impresa y el punto exacto, cita lo que dice el documento, muestra por qué no puede ser correcto, aporta la evidencia independiente y declara qué lectura implementa la biblioteca y qué test de regresión la fija. Un defecto listado aquí nunca es un defecto del método: en todos los casos la lectura pretendida se ha podido establecer a partir del propio documento o de la física.

Léela junto al informe de conformidad, que muestra los números que calcula la biblioteca; esta página explica el puñado de sitios donde lo que está mal es el valor esperado impreso.

El registro vive en docs/ERRATA.md y se trasplanta aquí en tiempo de compilación con make site-reports, así que ambos no pueden discrepar.

During the clean-room implementation of this library, every formula, constant and worked example is re-derived and recomputed independently from the source documents. That process occasionally surfaces defects in the sources themselves: misprints, worked examples that contradict their own normative text, and ambiguous wording. This file records each confirmed case with the evidence, what the library does about it, and whether it has been reported.

The registry covers every kind of published source the library implements from: standards (ISO, IEC, EN), guidance documents and technical reports (EASA, ECAC, NRL), textbooks and journal papers. Non-normative sources are marked as such in their entry.

Entries describe the specific printed editions cited. A defect listed here is not a defect of the method; in every case the intended reading could be established from the document itself or from physics, and the library implements that reading. Where the reading changes a number the library reports, the entry names the check or test that pins it; where the defect is a label, a cross-reference or a table the library never reads, the entry records that no change was required.

Status legend: unreported (recorded here only) / reported (submitted to the issuing body, with date and reference).

A claim that turns on the exact characters of a formula, constant, coefficient, symbol, inequality or table cell is verified against the page as printed, and its Evidence bullet cites that page by PDF page index and printed folio. Extracted text may locate a page; it is never quoted as “the print”, because PDF text layers delete glyphs silently (most of the sources cited here emit no √ at all, so f_T/√2 extracts as f_T/2). The page offset of each document is established empirically, because it differs per document and drifts between chapters of the same book. Entries that rest on something else, a recomputation or a comparison of two sentences, say so either in a leading notice or on the allowlist of scripts/check_errata_evidence.py, which is the check that enforces the rule; see CONTRIBUTING.md.

ISO 717-2:2020, Annex C, example C.1 (CI of the bare floor)

Location: Annex C, Table C.1 (printed p. 17) and the accompanying $C_{I}$ computation printed in the same cell.
The print: $L_{n, sum} = 83, 5238 \dots = 84 dB$ and $C_{I} = 84 - 15 - 79 = - 10 dB$ for the bare-floor example.
The problem: two independent defects in the same cell. (a) Clause A.2.1 defines $C_{I}$ from the energy sum over 100 Hz to 2500 Hz (the first fifteen one-third-octave bands); the printed value only reproduces if the 3150 Hz band is included, contradicting A.2.1. The correct sum over 100 Hz to 2500 Hz is 83,2613 dB, rounded 83, giving $C_{I} = - 11$ . (b) Even read as the sixteen-band sum the printed digits are wrong in the last place: the bare-floor $L_{n}$ column sums to 83,523 4 dB, not the printed 83,523 8 dB. The defect is confined to that cell, since the with-covering column of the same table prints $L_{n, sum} = 76, 0593 \dots$ and recomputes to 76,059 29 dB, reproducing every printed digit. Neither (a) nor (b) changes the rounded 84 dB, so only (a) moves $C_{I}$ .
Evidence: independent recomputation of both sums from the printed per-band levels (16 bands 83,523 38 dB, 15 bands 83,261 27 dB, with-covering 16 bands 76,059 29 dB); the 2013 edition of the same example prints $C_{I} = - 11$ . Verified on PDF page 23 (printed p. 17) and PDF page 17 (printed p. 11) of ISO 717-2:2020, and of PDF page 22 (printed p. 14) of ISO 717-2:2013.
Library behaviour: implements A.2.1 as written and pins $C_{I} = - 11$ with the 2013 print as the oracle (tests/reference_data/, conformance check “ISO 717-2 Annex C, Table C.1”).
Status: unreported.

ISO 717-2:2020, Annex C, example C.2 (covered floor: 800 Hz value and CI chain)

Location: Annex C, Table C.2 (printed p. 18), the $Δ L_{w}$ / $Δ L_{lin}$ worked example.
The print: (a) the 800 Hz reference-floor value is printed as 71,0 dB; (b) the $C_{I}$ line prints $L_{n, sum} = 75, 2527 \dots = 75 dB$ and $C_{I} = 75 - 15 - 63 = - 3 dB$ , feeding $Δ L_{lin} = 78 - 11 - (63 - 3) = 7 dB$ .
The problem: two independent defects. (a) The normative Table 4 reference floor is 71,5 dB at 800 Hz, and the column itself is a clean +0,5 dB per one-third octave ramp from 67,0 dB at 100 Hz to 72,0 dB at 1000 Hz, which the printed 71,0 dB breaks by repeating the 630 Hz cell. The misprint propagates along its own row and into the table’s total, three further cells the table prints and this entry previously did not name: the $L_{n, r, 0} - Δ L$ cell at 800 Hz is printed 64,0 dB ( $= 71, 0 - 7, 0$ ) where 71,5 gives 64,5; the unfavourable deviation is printed 3,0 dB ( $= 64, 0 - 61$ ) where the corrected cell gives 3,5; and the printed Sum 27,9 is the sum of the thirteen unfavourable deviations including that 3,0, where the corrected chain gives 28,4. None of it moves the rating: 28,4 dB is still below the 32,0 dB shift criterion, so $L_{n, w, r} = 63 dB$ and $Δ L_{w} = 15 dB$ either way. (b) The printed 75,2527 dB is exactly the energy sum of the wrong column over the wrong range: the measured floor “with covering” over all sixteen bands 100 Hz to 3150 Hz. A.2.1 defines $C_{I}$ from the reference floor with covering (the $L_{n, r, 0} - Δ L$ column) over 100 Hz to 2500 Hz (15 bands), which gives 75,674 dB (printed chain) or 75,710 dB (corrected 800 Hz cell), both round to 76 dB, so $C_{I, r} = 76 - 15 - 63 = - 2$ either way, giving $C_{I, Δ} = - 11 - (- 2) = - 9$ and $Δ L_{lin} = 6 dB$ , not the printed −3 / −8 / 7 dB chain.
Evidence: independent recomputation of every candidate sum and of every cell of the 800 Hz row from the printed per-band values; the printed 75,2527 reproduces to all printed digits only as the 16-band sum of the with-covering column, and every other cell of the $L_{n, r, 0} - Δ L$ and deviation columns reproduces exactly from the printed reference floor, so the 800 Hz row is the only one that does not. Verified on PDF page 24 (printed p. 18) and PDF page 13 (printed p. 7) of ISO 717-2:2020.
Library behaviour: derives the covered reference floor from the normative Table 4 values and sums per A.2.1, pinning $Δ L_{w} = 15 dB$ and $C_{I, Δ} = - 9$ ; the conformance check notes the provenance explicitly.
Status: unreported.

ISO 2631-5:2018, Annex C worked examples (male displayed formula, female R)

Location: Annex C: the displayed male worked example (82 kg male, $m_{z} = 0, 029 MPa / (m/s^{2})$ , printed p. 19) and NOTE 5 (64 kg female, $m_{z} = 0, 025 MPa / (m/s^{2})$ , printed p. 20).
The print: (a) the male example is displayed as
$R = {i = 0 \sum 20 - 1 [\frac{1 , 62 MPa ( 120 ) ^{1/6}}{6 , 75 MPa - 0 , 052 MPa ( 20 + i )}]^{6}}^{1/6} \approx 1, 22$
and (b) NOTE 5 states $R = 0, 97$ for the female case.
The problem: two independent defects. (a) The displayed male formula omits the $- S_{stat, i}$ term that normative Formula (C.3) puts in the denominator, and that the same annex fixes at $S_{stat} = 0, 029 \cdot 9, 81 = 0, 281 MPa$ in the sentence that follows the where-list of Formula (C.3). Evaluated exactly as displayed the sum gives $R = 1, 1497$ , which prints as 1,15, not the printed 1,22; restoring the missing term gives 1,2168 with the printed $S_{stat} = 0, 281 MPa$ and 1,2177 with the exact $m_{z} \cdot 9, 81 = 0, 2845 MPa$ , i.e. the printed 1,22 either way. The printed result is therefore right and the printed formula is not. (b) Exact recomputation of Formula (C.3) with NOTE 5’s own inputs ( $m_{z} = 0, 025$ , age coefficient 0,039, $b = 20$ , $n = 20$ , $N = 120$ ) gives $R = 0, 9621$ , which rounds to 0,96; the same code reproduces the male example exactly, and the note’s $S_{d} = 1, 40 MPa$ matches the exact 1,3992, so the discrepancy is confined to the last digit of the printed female $R$ .
Evidence: term-by-term recomputation of the C.3 sum under both readings of the denominator, with the male example as the discriminator: the printed 1,22 is reachable only with $- S_{stat}$ , and 1,15 only without it. Verified on PDF pages 23 (printed p. 17), 24 (printed p. 18), 25 (printed p. 19) and 26 (printed p. 20) of ISO 2631-5:2018.
Library behaviour: implements Formula (C.3) as written, with $- S_{stat}$ ; the male anchor pins 1,22 and the female test anchor keeps the printed 0,97 with a tolerance that documents the recomputed 0,9621.
Status: unreported.

EN 12354-1:2000 Formula (E.5) / ISO 12354-1:2017 E.3.4 (K24 clamp misprint)

Location: EN 12354-1:2000, Annex E, the wall-junction-with-flexible- interlayers block printed under Figure E.5 and numbered Formula (E.5) (printed p. 46), and ISO 12354-1:2017, E.3.4 NOTE 4. Annex E of the 2000 edition has only two numbered clauses, E.1 “Determination methods” and E.2 “Empirical data”, so “E.5” is a formula number, not a clause; an earlier revision of this entry cited it as a clause.
The print: $K_{24} = 3, 7 + 14, 1 M + 5, 7 M^{2} dB$ ; $0 \leq K_{24} \leq - 4 dB$ ; $0 dB / octave$ , i.e. the bound on the $K_{24}$ junction term is an empty interval; the 2017 edition repeats the 2000 misprint verbatim.
The problem: the interval is impossible as printed; the accompanying figure and the physics (the term is a reduction bounded below) indicate $- 4 dB \leq K_{24} \leq 0 dB$ .
Evidence: the Figure E.5 curve family on the same page runs the $K_{24}$ branch from 0 dB down to about −4 dB over the plotted mass ratios, which is the interval read in the other order. Verified on PDF page 48 (printed p. 46) of EN 12354-1:2000 and PDF page 52 (printed p. 46) of ISO 12354-1:2017.
Library behaviour: implements the clamp as $- 4 \leq K_{24} \leq 0$ with a misprint note in the docstring.
Status: unreported.

EN 12354-1:2000, Figure E.9 (E.7) (K24 stated in the figure-axis mass ratio)

Location: Annex E, Figure E.9 / Formula (E.7) (junction of lightweight double leaf wall and homogeneous elements), the $K_{24}$ line.
The print: $K_{24} = 3, 0 - 14, 1 M + 5, 7 M^{2} dB$ (for $m_{2} / m_{1} > 3$ ), under a figure whose x-axis is $m_{2} / m_{1}$ .
The problem: Annex E defines $M$ per transmission path as $M = l g (m_{⊥, i}^{'} / m_{i}^{'})$ (perpendicular element over the element carrying the path). The $K_{24}$ path 2→4 is carried by the homogeneous element ( $m_{2} = m_{4}$ ) with the leaf ( $m_{1}$ ) perpendicular, so the per-path $M$ is $lo g_{10} (m_{1} / m_{2})$ — but the printed $K_{24}$ line only matches its own figure’s curve when $M$ is read as the x-axis variable $lo g_{10} (m_{2} / m_{1})$ (e.g. −2,4 dB at $m_{2} / m_{1} = 3$ , −5,4 dB at 10). Read with the annex’s declared $M$ , the line contradicts the figure by $28, 2 \cdot ∣ lo g_{10} (m_{2} / m_{1}) ∣ dB$ . The same edition’s other $K_{24}$ line (Figure E.5, Formula (E.5)) does follow the declared per-path $M$ , so the two $K_{24}$ prints of the 2000 edition silently use different conventions. ISO 12354-1:2017 E.3.5 prints the relation consistently in the per-path convention of its Formula (E.3), $K_{24} = 3, 0 + 14, 1 M + 5, 7 M^{2}$ ; the two editions agree numerically (an earlier revision of this entry read the 2017 print as a sign misprint — re-derivation against both editions’ figures shows it is a convention recast, not a defect of the 2017 text).
Evidence: numerical evaluation of both forms against the Figure E.9 curve. Verified on PDF page 44 (printed p. 42), PDF page 48 (printed p. 46) and PDF page 50 (printed p. 48) of EN 12354-1:2000, and of PDF page 53 (printed p. 47) of ISO 12354-1:2017, whose E.3.5 prints its K24 line beside a Figure E.7 that carries no mass-ratio axis at all.
Library behaviour: implements the per-path convention uniformly (junction_vibration_reduction, mass_ratio = $m_{⊥, i}^{'} / m_{i}^{'}$ for every branch), so the E.7 double-leaf branch takes leaf-over-homogeneous ratios below 1/3 and evaluates $3, 0 + 14, 1 M + 5, 7 M^{2}$ .
Status: unreported.

EN 12354-2:2000, Formula (3) vs Annex E.3 (standardized impact level)

Location: Formula (3) and worked example E.3.
The print: Formula (3) defines $L_{n T}^{'} = L_{n}^{'} - 10 l g (0, 16 \cdot V / (A_{0} \cdot T_{0}))$ , which reduces exactly to $L_{n}^{'} - 10 l g (0, 032 \cdot V)$ , i.e. a reference volume of $31, 25 m^{3}$ . Annex E.3 states “from equation (3): $L_{n T, w}^{'} = L_{n, w}^{'} - 10 l g (V /30)$ ”.
The problem: the annex’s $V /30$ is a rounding of the formula’s own constant; the two differ by a constant 0,177 dB.
Evidence: direct algebra; both variants recomputed for the E.3 case (42,959 vs 42,782 dB, both rounding to 43 in that example). Verified on PDF page 7 (printed p. 5) and PDF page 34 (printed p. 32) of EN 12354-2:2000.
Library behaviour: implements the exact $0, 032 \cdot V$ form and documents the annex’s rounding.
Status: unreported.

EN 12354-3:2000, Formula (5) (reduced form of the normalized level difference)

Location: clause 3.1.5 “Relations between quantities”, Formula (5) (printed p. 6).
The print: $D_{2 m, n} = D_{2 m, n T} - 10 l g [0, 16 V / (T_{0} A_{0})] = D_{2 m, n T} - 10 l g 0, 32 V dB$ .
The problem: the reduced form is off by a factor of ten. Six lines above it, the where-list of clause 3.1.4 defines $A_{0}$ as “the reference equivalent sound absorption area, in square metres, for dwellings given as 10 m²”, and the where-list of clause 3.1.3 on the preceding page defines $T_{0}$ as “the reference reverberation time, in seconds, for dwellings given as 0,5 s”. So $0, 16/ (T_{0} A_{0}) = 0, 16/5 = 0, 032$ , not 0,32. Applied as printed, the reduced form shifts every normalized façade level difference by exactly $10 lo g_{10} 10 = 10 dB$ . The exact analogue in the companion part, EN 12354-2:2000 Formula (3), prints the same algebra correctly: $L_{n T}^{'} = L_{n}^{'} - 10 l g [0, 16 V / (A_{0} T_{0})] = L_{n}^{'} - 10 l g 0, 032 V dB$ . ISO 12354-3:2017 dropped the reduced form altogether: its Formula (5) prints only $D_{2 m, n} = D_{2 m, n T} - 10 l g [C_{sab} V / (A_{0} T_{0})]$ with $C_{sab} = 0, 16 s/m$ .
Evidence: direct algebra with the standard’s own $A_{0}$ and $T_{0}$ , and the side-by-side comparison with the correctly reduced Formula (3) of Part 2. Verified on PDF page 8 (printed p. 6) and PDF page 7 (printed p. 5) of EN 12354-3:2000, on PDF page 7 (printed p. 5) of EN 12354-2:2000 for its Formula (3), and on PDF page 12 (printed p. 6) of ISO 12354-3:2017 for the 2017 Formulae (4) and (5).
Library behaviour: unaffected. No code path implements the reduced form: the façade model computes $D_{2 m, n T}$ from Formula (13) (facade.py), and the survey method converts with the unreduced $D_{2 m, n} = D_{2 m} + k + 10 lo g_{10} [A_{0} T_{0} / (0, 16 V)]$ of ISO 10052 Clause 3.15 (survey_insulation.py). The two standardization constants that are pre-folded elsewhere in the library are both correct: $0, 032$ for the Part 2 impact form and $0, 32$ for the Part 1 airborne form $D_{n T} = R^{'} + 10 lo g_{10} (0, 16 V / (T_{0} S_{s}))$ , where the denominator is an area rather than $A_{0}$ .
Status: unreported.

EN 12354-3:2000, Formula (13) vs its own Annex F example (the “6” constant)

Location: clause 4.1, Formula (13) (printed p. 9), against the worked example of Annex F (printed pp. 27-28).
The print: Formula (13) gives $D_{2 m, n T} = R^{'} + Δ L_{fs} + 10 l g [V / (6 T_{0} S)] dB$ , while the Annex F.1.3 result table prints a $D_{2 m, n T}$ row that is exactly $R^{'} + 1, 5 dB$ in all five octave bands and in the single-number column (25,9/23,0/26,4/36,9/39,0 against 24,4/21,5/24,9/35,4/37,5, and 29,3 against 27,8).
The problem: on this constant the example is self-consistent and the formula is the outlier. (Two cells of the same annex table do not follow from its element rows, which is the subject of the next entry; the printed $+ 1, 5$ dB row holds in every band regardless, so the two defects are independent.) With the example’s own inputs ( $V = 50 m^{3}$ , $S = 11, 3 m^{2}$ , $T_{0} = 0, 5 s$ , $Δ L_{fs} = 0$ ), the Sabine form gives $10 lo g_{10} [0, 16 \cdot 50/ (0, 5 \cdot 11, 3)] = 1, 5104 dB$ , which is the printed +1,5 dB row; Formula (13) as printed gives $10 lo g_{10} [50/ (6 \cdot 0, 5 \cdot 11, 3)] = 1, 6877 dB$ . The gap is the constant: Formula (13)‘s “6” is a rounded $1/0, 16 = 6, 25$ , and $10 lo g_{10} (6, 25/6) = 0, 177 dB$ is exactly the discrepancy. ISO 12354-3:2017 replaced it with an explicit Sabine constant, printing Formula (4) as $D_{2 m, n T} = R^{'} + Δ L_{fs} + [10 l g (C_{sab} V / (T_{0} S))]$ with $C_{sab} = 0, 16 s/m$ , which is the constant the 2000 example already used. A previous revision of this entry attributed the 1,5 dB row to the example; the attribution is the other way round.
Evidence: evaluation of both constants against the printed Annex F rows, which agree with 0,16 to the 0,05 dB the table carries and disagree with the rounded 6 by a uniform 0,18 dB; and the 2017 recast, which adopts the example’s constant. The example’s single-number result $D_{2 m, n T, w} = 33 dB$ is insensitive to the difference and reproduces either way. Verified on PDF pages 11 (printed p. 9), 29 (printed p. 27) and 30 (printed p. 28) of EN 12354-3:2000, and of PDF page 12 (printed p. 6) of ISO 12354-3:2017.
Library behaviour: implements Formula (13) as printed, with the rounded 6; the test data records that the Annex F rows follow the exact 0,16 constant and sit 0,18 dB below the model.
Status: unreported.

EN 12354-3:2000, Annex F.1.3 (the 1 kHz and 2 kHz R’ cells)

Location: Annex F, table F.1.3 “Results for façade” (printed p. 28), the R' (equation 10) row.
The print: $R^{'}$ = 24,4 / 21,5 / 24,9 / 35,4 / 37,5 dB at 125 / 250 / 500 / 1000 / 2000 Hz.
The problem: the last two cells do not follow from the table’s own element rows. Formula (10), $R^{'} = - 10 lo g_{10} \sum τ_{e, i}$ , applied to the four $- 10 lo g_{10} τ_{e}$ columns printed immediately above gives 24,41 / 21,50 / 24,86 / 35,78 / 37,99 dB. The first three cells reproduce to the 0,05 dB the table carries; the 1 kHz and 2 kHz cells are printed 0,4 dB and 0,5 dB low.
Evidence: energy summation of the printed element rows band by band (1 kHz: 60,7 / 40,0 / 46,6 / 38,5 dB; 2 kHz: 66,7 / 41,0 / 43,6 / 44,5 dB). The $D_{2 m, n T}$ row below is a uniform $R^{'} + 1, 5 dB$ in every band including those two, so it inherits the same offset, and the single-number result $D_{2 m, n T, w} = 33 dB$ is insensitive to it and still reproduces. Verified on PDF page 30 (printed p. 28) of EN 12354-3:2000.
Library behaviour: the test data notes the inconsistency next to the affected anchor.
Status: unreported.

EN 12354-5:2009, Table F.1 and clause F.4.2 (reference force printed as 1 pN)

Location: Annex F, clause F.4.2: the symbol list of Formula (F.9), the sentence introducing the closed form, and the caption of Table F.1 (printed p. 59).
The print: ” $L_{F}$ is the force level in the source room, in dB re 1 pN”; ” $L_{F} = 10 l g 2, 5 f /1 0^{- 12}$ dB re 1 pN or $L_{F} = 10 l g 0, 8 f /1 0^{- 12}$ dB re 1 pN for one-third octave bands”; and “Table F.1 – Force level $L_{F}$ re 1 pN for the ISO tapping machine in octave bands”, whose eight cells read 139, 142, 145, 148, 151, 154, 156 and 156 dB.
The problem: the reference force of those levels is $1 0^{- 6}$ N, not 1 pN. Three independent readings agree, and none of them is compatible with the printed reference. (a) The annex’s own algebra. A power level re 1 pW built from a force level and a mobility is $L_{W} = L_{F} + 10 l g (F_{0}^{2} Y / W_{0})$ . Formula (D.5a) prints $L_{W s, c} = L_{F, e q} + 10 l g Y_{s}$ and Formula (D.9a) prints $L_{W s, c} = L_{F} - 5 - 10 l g f$ , which is the same expression evaluated at the mass-like source mobility $Y_{s} = (2 π f M)^{- 1}$ of a 0,5 kg tapping hammer. Neither carries a term for $F_{0}^{2} / W_{0}$ , so both balance only when $F_{0}^{2} / W_{0} = 1 s^{- 1}$ , that is $F_{0} = 1 0^{- 6}$ N; read re 1 pN each would fall 120 dB short of the level it defines. The velocity counterpart, Formula (D.10a), does print its reference term $10 l g (v_{ref}^{2} / W_{ref})$ and states the result cancels the $10 l g Z_{s}$ exactly, which it does at the $1 0^{- 9}$ m/s the standard itself gives as the velocity-level reference in clause F.4.2. The annex is therefore explicit and correct about the velocity reference and silent about the force one. (b) The machine that produces the table. The ISO tapping machine drops 0,5 kg hammers from 40 mm at ten impacts per second, so each impact transfers a momentum of 0,443 N·s and the force is a 10 Hz impulse train every harmonic of which carries 6,26 N r.m.s. Summing the harmonics that fall inside each octave band gives 139,4 / 142,4 / 145,4 / 148,4 / 151,4 / 154,4 dB re $1 0^{- 6}$ N from 31,5 Hz to 1 kHz, reproducing the first six cells of Table F.1 to within 0,5 dB; the 2 kHz and 4 kHz cells sit below that line, which is the roll-off the standard itself flags with “up till about 1000 Hz”. Re 1 pN the same cells would describe forces of tens of micronewtons, which no impact machine produces. (c) The companion standard. EN 15657:2018 Formula (15), which is where the structure-borne source data of Annex D comes from in the first place, writes the same force-to-power conversion “in dB re $F_{0} = 1 0^{- 6}$ N”, and $1 0^{- 6}$ N is the preferred reference force of ISO 1683.
Evidence: verified on PDF pages 61 and 62 (printed pp. 59 and 60) of BS EN 12354-5:2009, carrying clause F.4.2 with the symbol list of Formula (F.9), the closed form, the whole of Table F.1 and the symbol list of Formula (F.11) with its $1 0^{- 9}$ m/s velocity reference; and on PDF pages 45, 48 and 50 (printed pp. 43, 46 and 48) of the same edition, carrying Formulae (D.5a), (D.9a) and (D.10a).
Library behaviour: ships the printed cells unchanged and documents them re $1 0^{- 6}$ N. tapping_machine_force_level returns the eight values of Table F.1, tapping_machine_force_level_estimate the closed form and tapping_machine_characteristic_power_level Formula (D.9a) as printed; test_table_f1_is_referred_to_1e_6_newton_not_1_piconewton pins the reading against the mechanics of the machine.
Status: unreported.

EN 12354-5:2009, Figure D.3 Key (three curves under one symbol)

Location: Annex D, the Key of Figure D.3 (printed p. 47).
The print: three key rows, each labelled with the same symbol: $L_{W s, c, A} = 124 dB$ , $L_{W s, c, A} = 119 dB$ and $L_{W s, c, A} = 102 dB$ .
The problem: the figure’s own caption reads “Structure-borne sound power for the ISO-tapping machine: characteristic source power, installed power on a wooden floor and installed power on a concrete floor; the A-weighted power level is also indicated”. Only the first curve is a characteristic power; the other two are installed powers and their A-weighted totals are $L_{W s, inst, A}$ . The plotted curves settle the assignment: the first is flat at about 114,5 dB re 1 pW, which is the frequency-independent Formula (D.9a) result for the tapping machine, while the other two rise with frequency and lie below it, the concrete floor lowest, as $L_{W s, c} - D_{C, i}$ requires.
Evidence: verified on PDF page 49 (printed p. 47) of BS EN 12354-5:2009, the page carrying Figure D.3 with its Key and its caption.
Library behaviour: none required; no value is read from Figure D.3. test_formula_d9a_is_flat_at_about_115_db_per_third_octave pins the flat characteristic curve that the first key row belongs to.
Status: unreported.

ISO 12354-1:2017 Table L.3 / ISO 12354-2:2017 Table G.3 (perimeter sums)

Location: the input-data block below Table L.3 (printed p. 81) and the identical block below Table G.3 (printed p. 38), which lists the perimeter absorption sum $\sum l_{k} α_{k}$ of Formula (C.1) for the worked example.
The print: one value per element type: separating floor 2,364 m ( $S = 20 m^{2}$ ), external wall 2,375 m ( $S = 11 m^{2}$ ), internal wall 1,840 m ( $S = 13, 75 m^{2}$ ).
The problem: Formula (C.1) needs one sum per element, and the example has five elements with three different areas. Only two of the three printed values reproduce the columns they are supposed to drive: 2,375 m with $S = 11 m^{2}$ gives external wall 1 exactly, and 1,840 m with $S = 13, 75 m^{2}$ gives internal wall 2 exactly. The separating floor’s printed 2,364 m does not reproduce its own column at any band (0,074 9 against the printed 0,083 1 at 50 Hz, 0,026 4 against 0,029 0 at 500 Hz); 2,659 m does, at every band. The two elements with no printed value need 2,548 m (external wall 2, $S = 13, 75 m^{2}$ ) and 1,636 m (internal wall 1, $S = 11 m^{2}$ ).
Evidence: all five sums re-derived from Formula (C.4), $α_{k} = \sum_{j} f_{c, j} / f_{ref} 1 0^{- K_{ij} /10}$ , over the example’s own junction geometry with the unrounded Annex E indices: 2,659 / 2,375 / 2,548 / 1,636 / 1,839 m. The derivation returns the two printed values that are self-consistent with their own columns (2,375 m, and 1,839 m against the printed 1,840 m) and supplies the three that are missing or wrong, and every $η_{tot,situ}$ column of Table L.3 / G.3 then reproduces to $5 \cdot 1 0^{- 5}$ . The printed values applied to the wrong element of the same type miss by far more than that rounding: 2,375 m on external wall 2 gives 0,108 5 against the printed 0,114 9 at 50 Hz, and 1,840 m on internal wall 1 gives 0,085 0 against 0,077 0.
Library behaviour: in_situ_total_loss_factor takes $\sum l_{k} α_{k}$ as an input and perimeter_absorption_coefficient implements Formula (C.4); the Annex L fixture derives all five sums that way rather than using the printed block, and says so (tests/building/prediction/test_detailed_model.py).
Status: unreported.

ISO 12354-1:2017 Table L.3 / ISO 12354-2:2017 Table G.3 (external wall ηint)

Location: the same input-data block, external-wall line.
The print: $η_{int} = 0, 013$ for the 365 mm autoclaved aerated concrete external walls.
The problem: the example’s own element specification, and Annex B Table B.3 for autoclaved aerated concrete, give 0,012 5. Only 0,012 5 reproduces the tabulated $η_{tot,situ}$ : at 500 Hz Formula (C.1) gives $0, 0125 + 0, 00141 + 0, 03457 = 0, 0485$ , the printed value, where 0,013 would give 0,049 0.
Evidence: term-by-term recomputation of Formula (C.1) for both external walls at every band with each candidate $η_{int}$ .
Library behaviour: the Annex L fixture uses 0,012 5.
Status: unreported.

ISO 12354-1:2017, Table L.4 (second path block labelled 2d)

Location: Annex L, Table L.4 (printed p. 82), the right-hand block headed “Transmission path 2d”.
The print: the block gives $α_{i, situ}$ = 6,3 to 14,1, $D_{v, ij, situ}$ = 11,0 to 13,6 and $R_{ij}$ = 43,9 to 84,6 dB.
The problem: those are the numbers of path 4d (internal wall 2 to the separating floor), not of path 2d (external wall 2). Table L.1 of the same annex prints the whole $R_{4 d}$ column, 43,9 to 84,6 dB, and the block’s $R_{ij}$ column is that column cell for cell. What settles it band by band is the other two columns, which cannot be confused: external wall 2 has $α_{i, situ} = 10, 3 m$ at 50 Hz ( $S = 13, 75 m^{2}$ , $η_{tot} = 0, 1149$ ) while internal wall 2 has 6,3 m ( $η_{tot} = 0, 0703$ ), the printed value; and $D_{v, ij, situ}$ follows the floor-to-internal-wall $K_{ij}$ of 8,8 dB, which gives 11,0 to 13,6 dB, not the floor-to-external-wall 6,4 dB, which gives 9,6 to 11,9 dB.
Evidence: independent recomputation of Formulae (10), (11) and (15) for both candidate paths at every band. Path 4d reproduces all three columns of the block, $α_{i, situ}$ to 0,05 m and $D_{v, ij, situ}$ and $R_{ij}$ to 0,05 dB, which is the printed resolution. Path 2d departs from the block’s $R_{ij}$ column by 0,1 dB to 7,0 dB depending on the band, and comes closest between 100 Hz and 160 Hz (0,5 / 0,5 / 0,1 dB), so $R_{ij}$ alone does not identify the path over those bands; $α_{i, situ}$ (10,3 against 6,3 m at 50 Hz) and $D_{v, ij, situ}$ (1,4 dB to 1,7 dB apart in every band) do.
Library behaviour: the test that asserts the block builds it as path 4d and names the mislabelling.
Status: unreported.

ISO 12354-1:2017, Table L.1 (non-integer weighted ratings)

Location: Annex L, Table L.1 (printed p. 79), the $R_{w}$ row and the sentence below it, and the corresponding $L_{n, w}$ row of ISO 12354-2:2017 Table G.1.
The print: the $R_{w}$ row gives one decimal for every path (75,1 / 84,5 / 70,6 / … and 57,8 in the total column) while the sentence immediately below states $R_{w}^{'} (C; C_{tr}) = 57, 9 (- 2; - 8) dB$ .
The problem: ISO 717-1 rates by shifting the reference curve in 1 dB steps, so a weighted rating is an integer; the printed one-decimal values are the reference curve shifted continuously until the sum of unfavourable deviations equals exactly 32,0 dB. The airborne $R_{w}$ row of Table L.1 truncates that continuous value to one decimal while the sentence below it rounds, which is why the same quantity appears twice as 57,8 and 57,9; the impact $L_{n, w}$ row of Table G.1 rounds instead (29,58 prints as 29,6 and 40,98 as 41,0), so the truncation is a property of the airborne row only. The spectrum adaptation terms inherit the offset: with the ISO 717-1 rating of 57 dB they are $C = - 1$ and $C_{tr} = - 7$ , and the printed (−2 ; −8) is exactly the pair shifted by the same 0,86 dB.
Evidence: a continuous-shift solve of the ISO 717-1 reference curve against the printed per-band spectra reproduces every printed value in both rows ( $R_{D d}$ 75,12 against 75,1; $R_{D 1}$ 84,54 against 84,5; $R_{11}$ 70,66 against 70,6; the total 57,86 against 57,8 / 57,9; on the impact side $L_{n, D f 1}$ 29,58 against 29,6 and the total 40,98 against 41,0), whereas the ISO 717-1 1 dB-step ratings of the same spectra are 75, 84, 70 and 57 dB. Verified on PDF page 85 (printed p. 79) of ISO 12354-1:2017.
Library behaviour: weighted_rating / weighted_impact_rating implement ISO 717-1/-2 as written, so the detailed model returns $R_{w}^{'} = 57 dB$ and $L_{n, w}^{'} = 41 dB$ ( $C_{I} = 2$ ) for the example; the test pins those and documents the printed values.
Status: unreported.

ISO 12354-2:2017, Table G.1 (50 Hz to 80 Hz flanking columns)

Location: Annex G, Table G.1 (printed p. 36), the four $L_{n, D f}$ columns, 50 Hz, 63 Hz and 80 Hz rows.
The print: $L_{n, D f 1}$ = 47,3 / 44,9 / 46,2 dB.
The problem: Table G.4 of the same annex prints the same path Df for external wall 1, from the same inputs, as 47,8 / 45,9 / 47,0 dB. The two tables cannot both be right, and from 100 Hz upwards they agree exactly.
Evidence: Formula (12) evaluated from the annex’s own Table G.3 columns ( $L_{n, situ}$ , $R_{situ}$ ) and the Table G.4 $D_{v, ij, situ}$ and $Δ L_{situ}$ columns gives 47,80 / 45,85 / 46,95 dB, reproducing the printed 47,8 / 45,9 / 47,0 of Table G.4 to 0,05 dB and Table G.1 only from 100 Hz upwards. Carrying the same recomputation through the whole chain puts external wall 2 low by 0,5 dB to 1,0 dB over the same three bands and the two internal walls low by up to 0,5 dB at 50 Hz and 63 Hz (their 80 Hz cells agree). From 100 Hz upwards no flanking column deviates by more than 0,15 dB. Correcting the affected cells raises the printed total $L_{n}^{'}$ only slightly: 58,6 to 58,7 dB at 50 Hz, 57,0 to 57,2 dB at 63 Hz, 55,9 to 56,1 dB at 80 Hz.
Library behaviour: the test asserts Table G.4 in full, the Table G.1 direct column over the whole range, and the Table G.1 flanking columns from 100 Hz upwards, naming the disagreement.
Status: unreported.

ISO 12354-2:2017, Table G.8 (junction Kij and m’i)

Location: Annex G, Table G.8 (printed p. 40), the internal wall to external wall rigid T junction.
The print: row “Int. wall 1/2 - Ext. wall 1/2” gives $K_{ij} = 6, 6 dB$ ; the row below it, “Ext. wall 1/2 - Ext. wall 1/2”, gives $m_{i}^{'} = 2, 19 kg/m^{2}$ .
The problem: two independent misprints. The rigid-T corner branch $K_{12} = 5, 7 + 5, 7 M^{2}$ with $M = lo g_{10} (360/219) = 0, 2156$ gives 5,97, i.e. 6,0, and ISO 12354-1:2017 Table L.8 prints 6,0 for the identical junction of the identical example. And the external wall’s mass per unit area is $219, 0 kg/m^{2}$ throughout the example, not 2,19 (a factor 100).
Evidence: Annex E evaluation of the corner branch; the same table’s own other rows and the whole of ISO 12354-1 Annex L use $219, 0 kg/m^{2}$ . Verified on PDF page 46 (printed p. 40) of ISO 12354-2:2017, whose Table G.8 mass columns are headed m'i and m'orthogonal, and PDF page 89 (printed p. 83) of ISO 12354-1:2017.
Library behaviour: uses 6,0 dB and $219, 0 kg/m^{2}$ .
Status: unreported.

ISO 12354-2:2017, Table G.6 (mislabelled row)

Location: Annex G, Table G.6 (printed p. 40), internal wall to separating floor rigid cross junction.
The print: a row labelled “Ext. wall 1/2 – Int. wall 1/2” with m'i = 360,0, m'orthogonal = 484,0 and $K_{ij} = 11, 0 dB$ .
The problem: Table G.6 describes the internal wall to separating floor cross junction; no external wall meets it. The masses and the value are those of the in-line internal-wall path, and ISO 12354-1:2017 Table L.6 prints the same row correctly as “Int. wall 1/2 - Int. wall 1/2”.
Evidence: the rigid-cross through branch $8, 7 + 17, 1 M + 5, 7 M^{2}$ with $M = lo g_{10} (484/360)$ gives 10,99, the printed 11,0, for the internal wall. Verified on PDF page 46 (printed p. 40) of ISO 12354-2:2017 and PDF page 89 (printed p. 83) of ISO 12354-1:2017.
Library behaviour: treats the row as the internal-wall in-line path.
Status: unreported.

ISO 12354-1:2017 Table L.10 / ISO 12354-2:2017 Table G.10 (element label)

Location: the simplified-model input table of both parts, fourth row: Table L.10 (printed p. 84) and Table G.10 (printed p. 41).
The print: ISO 12354-1 prints “Internal wall 4 (F = f = 4)”; ISO 12354-2 prints “Internal wall 4 (f4)”: the two parts label the row differently, and an earlier revision of this entry quoted the Part 1 form for both.
The problem: the example has two internal walls; the element indexed $F = f = 4$ is internal wall 2 ( $5, 00 m \times 2, 75 m$ , $S = 13, 75 m^{2}$ ), as the detailed-model tables of the same annexes label it.
Evidence: the row’s own $S = 13, 75 m^{2}$ and $l_{ij} = 5, 0 m$ match internal wall 2 of Table L.1 / G.1. Verified on PDF page 90 (printed p. 84) of ISO 12354-1:2017 and of PDF page 47 (printed p. 41) of ISO 12354-2:2017, with the detailed-model column labels read on PDF page 85 (printed p. 79) of ISO 12354-1:2017 and of PDF page 42 (printed p. 36) of ISO 12354-2:2017.
Library behaviour: none needed; the numbers are unaffected.
Status: unreported.

ISO 12354-1:2017, Table D.1 (1 600 Hz covered by two rows)

Location: Annex D, Table D.1 (printed p. 39), which reads the weighted sound reduction index improvement of an interior lining off its resonance frequency.
The print: the last two rows are “630 to 1 600 -> -10” and “1 600 <= f0 <= 5 000 -> -5”.
The problem: 1 600 Hz belongs to both rows, with different values, and Clause D.2.2 requires $f_{0}$ to be “rounded to the centre frequency of the one-third-octave band in which fo falls”, so 1 600 Hz is a value the table is actually read at rather than an unreachable edge. Because the rounding is mandatory, the ambiguity is not a single point: every raw resonance frequency in the 1 600 Hz band, that is from 1 412,5 Hz to 1 778,3 Hz (ISO 266 band edges), lands on it. Every other boundary in the table is a distinct band centre (200, 250, 315, 400, 500 Hz), and no other pair of rows overlaps.
Evidence: the printed table itself, on PDF page 45 (printed p. 39) of ISO 12354-1:2017: the two rows are separately ruled and share the endpoint verbatim, “630 to 1 600” and “1 600 <= f0 <= 5 000”. Neither row can be discarded, because 630 Hz to 1 250 Hz has no other entry and 2 000 Hz to 5 000 Hz has none either. The predecessor edition gives the earlier, unambiguous reading: EN 12354-1:2000 Table D.3, verified on PDF page 43 (printed p. 41) of that edition, prints the same pair of rows as “630 - 1 600 -> -10” and ”> 1 600 -> -5”, strictly greater, so in 2000 exactly 1 600 Hz took -10 dB with nothing to decide. The 2017 rewrite replaced ”> 1 600” with “1 600 <= f0 <= 5 000” while leaving “630 to 1 600” untouched, which is what creates the overlap; what the rewrite intended at the shared endpoint the text does not say.
Library behaviour: weighted_lining_improvement returns the more conservative -10 dB at exactly 1 600 Hz and -5 dB above it, the 2000 reading, with the ambiguity named in the docstring and pinned in tests/building/prediction/test_resilient_layers.py.
Status: unreported.
Related, not an erratum: NOTE 1 of the same table sets a floor of 0 dB on the 30 Hz to 160 Hz branch $74, 4 - 20 lo g_{10} (f_{0}) - R_{w} /2$ . Inside the validity box Clause D.2.2 states for the table ( $30 Hz \leq f_{0} \leq 160 Hz$ , $20 dB \leq R_{w} \leq 60 dB$ ) the branch never reaches it: its minimum is $74, 4 - 20 lo g_{10} (160) - 60/2 = 0, 32 dB$ . The floor is therefore inactive for every input the table is stated for, but it was not always: the 2000 edition tabulated the low branch as four discrete rows ending in “160 -> 28 - Rw/2”, whose minimum is $28 - 60/2 = - 2 dB$ , so NOTE 1 was operative there. The 2017 continuous fit sits 2,3 dB above it at that corner and left the note vestigial. The library keeps the floor because the note is still printed.

ISO 15186-1, Clause 3.9, Formula (8) (sign of the 10 lg N term)

Location: Clause 3.9, Formula (8) (printed p. 3), the intensity element normalized level difference for N small building elements measured together. The print read here is BS EN ISO 15186-1:2003, the identical-text British adoption; the entry previously carried the heading ":2000", the year of the ISO edition the library’s docstrings cite, which is not the copy that was read.
The print: $D_{I, n, e} = L_{p 1} - 6 - (L_{I n} + 10 l g (S_{m} / A_{0}) + 10 l g (N))$ , i.e. the $10 l g N$ term is subtracted.
The problem: the subtracted sign cannot be derived. Measuring $N$ identical units within one measurement surface raises the transmitted power (and hence $L_{I n} + 10 lo g_{10} S_{m}$ ) by $10 lo g_{10} N$ , so recovering the per-unit $D_{I, n, e}$ requires adding $10 lo g_{10} N$ . The pressure-based equivalent, ISO 10140-2:2010 Formula (6), prints exactly that correction ( $D_{n, e} = L_{1} - L_{2} + 10 lo g_{10} (n A_{0} / A)$ ), and ISO 15186-2:2010 Formula (12) prints Formula (8) without any $N$ term (the $N = 1$ case, with which both signs agree). As printed, installing more units would lower the per-unit rating by $20 lo g_{10} N$ relative to the derivable value.
Evidence: derivation from the diffuse-field receiving-room relation $L_{2} = L_{W} + 10 lo g_{10} (4/ A)$ against ISO 10140-2:2010 Formula (6); cross-check against ISO 15186-2:2010 Formula (12) and Hopkins, Sound Insulation (2007), Eq. 3.45. Verified on PDF page 11 (printed p. 3) of BS EN ISO 15186-1:2003, with the cross-check read on PDF page 11 (printed p. 11) of ISO 10140-2:2010.
Library behaviour: implements the derivable per-unit form (intensity_element_normalized_difference, $+ 10 lo g_{10} N$ ) and emits a warning whenever $n > 1$ , where the result deviates from the print.
Status: unreported.

ISO 10848-1:2006, Clause 8.1.1, Formula (20) (spurious π in the critical frequency)

Location: Clause 8.1.1, Formula (20), the thin-plate critical frequency used by the test-facility flanking criterion of Formula (19).
The print: $f_{c} = c_{0}^{2} / (1, 8 c_{L} \cdot h \cdot π)$ .
The problem: the constant 1,8 is itself the rounded $2 π / 12 \approx 1, 814$ of the thin-plate dispersion relation, so the extra $π$ double-counts it and would misplace $f_{c}$ by a factor $π$ (e.g. a 100 mm concrete element with $c_{L} = 3500 m/s$ : 187 Hz without the $π$ , 59 Hz with it, far from any measured coincidence dip).
Evidence: derivation from the thin-plate dispersion relation (Hopkins, Sound Insulation (2007), Eq. 2.201, $f_{c} = c_{0}^{2} / (1, 8 c_{L} h)$ ); ISO 12354-1:2017 prints the same $π$ -free form in its symbol definitions ( $f_{c} = c_{0}^{2} / (1, 8 c_{L} t)$ ).
Library behaviour: implements the $π$ -free form (phonometry.building.measurement.flanking_transmission.critical_frequency), with a misprint note in the docstring.
Status: corrected upstream — ISO 10848-1:2017 (second edition) prints the $π$ -free form in its Formula (5), $f_{c} = c_{0}^{2} / (1, 8 h c_{L})$ , confirming the 2006 print as a misprint. No report is needed. The entry is retained because the library cites the 2006 edition, whose print carries the defect; the 2017 edition stands as the confirmation.

UNE-EN 15657:2018, Clause 7.1, Formula (14) (reference mass dimensionally inconsistent with the quantity it normalises)

Location: Clause 7.1, the sentence introducing Formula (14) (printed p. 14) and Formula (14) itself (printed p. 15), the structural power level injected into the reception plate.
The print: the sentence reads “a partir del nivel de velocidad promediado espacialmente de la placa $L_{v}$ , de la masa por unidad de superficie $m$ , del área de la placa $S$ y del factor de pérdida $η$ , utilizando $f_{0} = 1$ Hz, $m_{0} = 1$ kg y $S_{0} = 1$ m² como referencias”, above $L_{W s} = (10 l g (\frac{2 π f m η S}{f _{0} \cdot m _{0} \cdot S _{0}})) dB + L_{v} - 60 dB$ .
The problem: the same sentence defines $m$ as a mass per unit area, in kg/m², and its reference $m_{0}$ as 1 kg. With $m$ in kg/m² and $S$ in m², the group $2 π f η m S / (f_{0} m_{0} S_{0})$ is dimensionless only if $m_{0}$ is 1 kg/m²; as printed it carries a leftover m⁻². The closing constant confirms the intended reading: $10 l g (f_{0} m_{0} S_{0} v_{0}^{2} / P_{0}) = - 60$ dB with $v_{0} = 1 0^{- 9}$ m/s and $P_{0} = 1$ pW closes in watts only when $f_{0} m_{0} S_{0}$ has the units of an area density times an area times a frequency. The numeric result is unaffected, because $10 l g (1) = 0$ whichever unit is attached, which is why the slip survives a worked example.
Evidence: dimensional analysis of Formula (14) against the definition of $m$ in the sentence above it, and against the $- 60$ dB constant it closes on; the sentence and the formula were read as images, not from extracted text. Verified on PDF page 14 (printed p. 14) and PDF page 15 (printed p. 15) of UNE-EN 15657:2018. Only the Spanish-language adoption was read, so this entry does not establish whether the English EN 15657:2018 print carries the same reference.
Library behaviour: no change required. characteristic_reception_plate_power takes mass_per_area in kg/m² and reproduces the standard’s own worked values, so the intended reading is the implemented one; the guide and the docstring keep the printed reference and name this entry beside it.
Status: unreported.

ISO 12999-1:2020, Table 4 (missing 500 Hz row)

Location: Table 4 (in-situ uncertainties per band).
The print: the 2020 edition’s table omits the 500 Hz row that the 2014 edition prints (situation B 1,2 dB / situation C 0,8 dB).
The problem: likely an editorial omission; the surrounding rows are unchanged between editions and the text does not mention removing the band.
Evidence: side-by-side comparison of the 2014 and 2020 prints.
Library behaviour: follows the 2020 print as published, with the omission documented in the module.
Status: unreported.

ISO 12999-2:2020, Clause 8 wording vs Tables 4 and 5

Location: Clause 8 “Reporting uncertainties” (printed pp. 5-6), the where-list under Formula (10), against the worked Tables 4 and 5 (printed p. 7). An earlier revision of this entry called the clause “expression of results”, which is not its printed title.
The print: the where-list defines $u$ as “the standard uncertainty determined in accordance with Clause 5, Clause 6 or Clause 7 rounded to two decimal digits for absorption coefficients or one decimal digit for all other quantities”, and Formula (10) then forms $U = k \cdot u$ .
The problem: the document’s own Tables 4 and 5 only reproduce when $U$ is computed from the unrounded $u$ and rounded last. Neither table prints a $u$ column at all (each has only the coefficient $α_{s}$ or $α_{p}$ and $\pm U$ with $k = 2$ ), so the printed $U$ values are the whole of the evidence, and 11 of the 25 are unreachable under the literal clause wording.
Evidence: recomputation of all 25 entries (Table 4: 20 rows, Table 5: 5 rows) from Formula (1) with the Table 1 constants and from Formula (4) with the Table 2 constants, under both conventions. Round-last reproduces 25 of 25; round-first misses 11 of 25 (63, 125, 160, 200, 250, 1250, 1600, 2000, 3150 and 4000 Hz of Table 4, and 250 Hz of Table 5). An earlier revision of this entry quoted the count as “10 of 20”, which is neither the right numerator nor the right number of entries. Verified on PDF pages 9 (printed p. 3), 10 (printed p. 4), 11 (printed p. 5) and 13 (printed p. 7) of ISO 12999-2:2020.
Library behaviour: rounds last, matching the tables; the convention is documented and tested.
Status: unreported.

ISO 12999-2:2020, Table 5 (octave-band data under a one-third-octave header)

Location: clause 8, Table 5 “Example for the practical sound absorption coefficient, αp, and its expanded uncertainty under reproducibility conditions” (printed p. 7).
The print: the frequency column of Table 5 is headed “One-third octave midband frequency / Hz” and its rows are 250, 500, 1 000, 2 000 and 4 000 Hz.
The problem: those five frequencies are the octave-band series of ISO 11654, which is what the practical sound absorption coefficient $α_{p}$ is defined over; they are not a one-third-octave series, and no one-third octave band is missing between them. The document contradicts itself on the same quantity two pages earlier: Table 2, which supplies the $m$ and $n$ constants of Formula (4) for exactly these five frequencies, is headed “Octave midband frequency”. The same header text stands over Table 4 on the same page, where it is correct: that table carries a genuine one-third-octave series, 63 Hz to 5 000 Hz in 20 rows.
Evidence: the five tabulated frequencies themselves, and the “Octave midband frequency” header of Table 2 for the same $α_{p}$ constants. Verified on PDF page 13 (printed p. 7) and PDF page 11 (printed p. 5) of ISO 12999-2:2020.
Library behaviour: _TABLE2 in uncertainty.py is keyed by octave midband frequency, following Table 2 and the ISO 11654 definition of $α_{p}$ rather than the Table 5 header.
Status: unreported.

ISO 10052:2021, Table 4 volume-range header

Location: Table 4 (reverberation-index estimator), volume-range header.
The print: the header reads “60 ≤ V < 150” while the body text says the method applies to rooms “up to 150 m³”.
The problem: the boundary $V = 150 m^{3}$ is included by the text and excluded by the header.
Evidence: direct comparison of header and clause text.
Library behaviour: accepts $V = 150$ (follows the text), with the ambiguity noted.
Status: unreported.

ISO 17208-2:2019, Clause 5 uncertainty band coverage

Location: Clause 5 (representative expanded uncertainties), printed p. 4.
The print: “5 dB for the low frequency (10 Hz to 100 Hz) bands, 3 dB for the mid frequency (125 Hz to 16 000 Hz) bands, and 4 dB for the high frequency (>20 000 Hz) bands”.
The problem: the 20 kHz one-third-octave band itself is left unassigned: the mid range ends at 16 kHz inclusive and the high range starts strictly above 20 kHz. ISO 17208-1:2016, from which clause 5 says the values are taken, prints the same three ranges with “≥20 000 Hz”, which closes the gap; Part 2 degraded the $\geq$ to a $>$ . The 20 kHz band is not a corner case for this document: ISO 17208-1 Table 1 requires the measurement to cover “20 000 Hz (minimum)” as its upper one-third-octave band. An earlier revision of this entry said “nothing covers 16 kHz to 20 kHz inclusive”, which is wrong at the lower end: 16 kHz is covered.
Evidence: the two clauses side by side. Verified on PDF page 10 (printed p. 4) of ISO 17208-2:2019 and PDF page 22 (printed p. 16) of ISO 17208-1:2016.
Library behaviour: applies the conservative 4 dB high-band value from the 20 kHz band upwards, following Part 1, with the gap documented.
Status: unreported.

ECMA-418-1:2024 (3rd edition), clause 4.1.1 NOTE 2 (upper limit of the discrete-tone range)

Location: clause 4.1.1 “frequency range of interest”, NOTE 2 (printed p. 2). An earlier revision of this entry cited clause 4.1.2, which is the definition of “ITT equipment” and says nothing about frequency.
The print: “From viewpoint of test implementation by using FFT analyser, the frequency range of discrete tones are between 89,1 Hz and 11 220 Hz inclusive, referred to the discrete tone frequency range of interest.”
The problem: every formula and table of the standard works to 11 200 Hz: the Table 2 and Table 3 band-edge fits are stated for $11200 \geq f_{t} > 1600$ , and clauses 10, 12.3 and 12.4 permit FFT data with $f_{1} < 89, 1 Hz$ and $f_{2} > 11200 Hz$ . The two numbers are the same quantity to different precision rather than a typographical error: $10000 \cdot 2^{1/6} = 11224, 6 Hz$ is the upper edge of the 10 kHz one-third-octave band that closes the range of interest, which rounds to 11 220 Hz at four significant figures and to 11 200 Hz at three. An earlier revision of this entry called it a typo and added that “no other clause mentions 11 220 Hz”; the last x-axis tick of Figure 6 (printed p. 20) is labelled 11220. What clause 4.1 does carry is a structural defect: 4.1.2 “ITT equipment” repeats 4.1.1’s NOTE 1 verbatim (“This range was selected to be identical to that of ECMA-74:2022, 3.1.3”), although 4.1.2 defines no range at all, and clause 10 then cross-references “NOTE 1 of 4.1.2” for the discrete-tone range, which is the duplicated note rather than the NOTE 2 that states it.
Evidence: the arithmetic above, and the Table 2/3 ranges and Figure 6 axis read side by side with NOTE 2. Verified on PDF page 10 (printed p. 2), PDF page 18 (printed p. 10), PDF page 25 (printed p. 17) and PDF page 28 (printed p. 20) of ECMA-418-1:2024 (3rd edition).
Library behaviour: uses the internally consistent $89, 1 Hz$ to 11 200 Hz range (upper end exclusive per the formulas), with a code note in tonality.py.
Status: unreported.

ECMA-418-1:2024 (3rd edition), Formula (21) (repeated constant term)

Location: clause 12.3, Formula (21) (printed p. 17), the curve fit for the lower band-edge frequency $f_{1, L}$ of the lower critical band.
The print: $f_{1, L} = C_{L, 0} + C_{L, 0} f_{t} + C_{L, 2} f_{t}^{2}$ .
The problem: the linear coefficient repeats the constant term. The where-list immediately below the formula declares ” $C_{L, 0}$ , $C_{L, 1}$ , $C_{L, 2}$ are constants given in Table 2”, Table 2 tabulates a $C_{L, 1}$ column, and the parallel Formula (22) for the upper band edge prints $f_{2, U} = C_{U, 0} + C_{U, 1} f_{t} + C_{U, 2} f_{t}^{2}$ correctly. The misprint is numerically fatal, not cosmetic: over the middle fit range ( $171, 4 \leq f_{t} \leq 1600$ ) Table 2 gives $C_{L, 0} = - 149, 5$ and $C_{L, 1} = 1, 001$ , so the printed form returns $- 149, 5 - 149, 5 f_{t} - 6, 90 \cdot 1 0^{- 5} f_{t}^{2}$ , negative everywhere, instead of a band edge a little below $f_{t}$ .
Evidence: the formula, its own where-list and Table 2 on one page, with Formula (22) as the consistent control. Verified on PDF page 25 (printed p. 17) of ECMA-418-1:2024 (3rd edition).
Library behaviour: implements the $C_{L, 1}$ reading, which is the only one that returns a usable band edge, with a code note in tonality.py.
Status: unreported.

ECMA-418-1:2024 (3rd edition), clause 11.3 (unresolved field references)

Location: clause 11.3 “Determination of masking noise level” (printed p. 12), the sentence introducing the critical bandwidth.
The print: “The critical bandwidth Δf_c is determined from Formula Error! Reference source not found.Error! Reference source not found. with f_0 set equal to the frequency of the discrete tone under investigation, f_t”.
The problem: two unresolved word-processor field references were typeset, in bold, in place of the formula numbers, and shipped in the published third edition. The intended targets are unambiguous from the rest of the sentence, which goes on to name Formulae (4) and (5) or (7) and (8) for the band edges: the critical bandwidth itself is Formula (2), and Formula (3) is the relation $f_{2} - f_{1} = Δ f_{c}$ that turns it into band edges.
Evidence: the clause as printed. Verified on PDF page 20 (printed p. 12), PDF page 18 (printed p. 10) and PDF page 30 (printed p. 22) of ECMA-418-1:2024 (3rd edition).
Library behaviour: none required; the library implements the critical bandwidth from Formulae (3)/(6) directly.
Status: unreported.

ECMA-418-2:2025 (4th edition), clause 5.1.5.2 (last block index)

Location: clause 5.1.5.2, the segmentation of the zero-padded signal for the roughness/fluctuation-strength block sizes.
The print: the index of the last block is given as $l_{last} = ⌈(n + s_{b}) / s_{h} ⌉$ .
The problem: the formula is internally inconsistent: blocks placed at that index overrun the zero-padded signal defined by clause 5.1.2.2, and the resulting Formula (103) time grid becomes non-monotonic. The only self-consistent reading is to stop at the last block that fits inside the padded signal and align it flush with its end.
Evidence: direct evaluation of the block start indices against the padded length for the clause 7.1.1 block/hop sizes; the flush-to-end reading reproduces the Clause 7 roughness calibration ( $1 asper$ ) to $0, 9999$ .
Library behaviour: implements the flush-to-end reading with a code note in roughness_ecma.py.
Status: unreported.

ECMA-418-2:2025 (4th edition), clause 9.1.4, Formula (127) (HSA kernel phase)

Location: clause 9.1.4, Formula (127), the spectral kernel of the envelope analysis window used by the High-resolution Spectral Analysis.
The print: the kernel’s phase factor is $exp (- j \cdot 2 π \cdot f_{n} (k) \cdot (\tilde{s}_{b} - n_{z e} + n_{z b} - 1))$ .
The problem: the kernel is, by construction, the DFT of the rectangular analysis window of Formula (120) modulated to the candidate rate; that is the model Formula (124) fits to the measured DFT spectrum. That DFT has the phase $exp (- j \cdot π \cdot f_{n} \cdot (\tilde{s}_{b} - n_{z e} + n_{z b} - 1))$ ; the printed factor doubles it (and is also inconsistent with the $π$ arguments of the printed sine terms of the same formula). With the printed phase the fitted model cannot reproduce the spectrum of a noiseless windowed sinusoid, contradicting the clause’s own statement that the HSA achieves “theoretically infinite resolution for signals without noise”.
Evidence: independent derivation of the window DFT plus numerical recomputation: with $π$ the least-squares fit recovers the constant part, amplitudes and phases of synthetic noiseless envelopes to machine precision and the Formula (135) residual vanishes; with the printed $2 π$ the kernel deviates from the window DFT by amounts of the order of the kernel itself and the residual stays of the order of the signal energy.
Library behaviour: implements the $π$ reading, pinned by a regression test on the exact recovery of synthetic line pairs.
Status: unreported.

ECMA-418-2:2025 (4th edition), clause 9.1.5, Formula (144) (bin offset)

Location: clause 9.1.5, Formula (144), the modulation rate of a local maximum of the envelope power spectrum.
The print: the rate is the three-bin amplitude-weighted centroid of the peak position minus one, scaled by $Δ f$ .
The problem: clause 9.1.4 (below Formula (122)) defines the spectral index $k$ as mapping to the modulation rate $k \cdot \tilde{r}_{s} / \tilde{s}_{b}$ with $k$ starting at 0. A symmetric local maximum at bin $k$ has centroid $k$ , and the printed formula then assigns it the rate $(k - 1) \cdot Δ f$ , one full bin ( $0, 73 Hz$ ) low, which at fluctuation-strength rates is fatal (a true $1, 46 Hz$ modulation would be reported as $0, 73 Hz$ ). The offset is only consistent with 1-based spectral-line positions, contradicting the standard’s own definition of $k$ .
Evidence: cross-check of Formula (144) against the $k$ -to-rate mapping stated below Formula (122).
Library behaviour: uses the centroid directly (no offset) with the 0-based $k$ of Formula (122).
Status: unreported.

ECMA-418-2:2025 (4th edition), clause 9.1.7 (units of the fine-tuning constants)

Location: clause 9.1.7, Formulae (149)-(152), the damped Newton fine tuning of the dominant modulation rate.
The print: differential step $Δ x = 1 0^{- 5}$ , damped-step cap $2 \cdot 1 0^{- 4}$ , stop tolerance $1 0^{- 7}$ and an iteration limit of 40, with the starting point $x_{0} = \tilde{f}_{c, i_{max}}$ (a rate in Hz) and the failure check $∣ f_{c, 1, opt} - \tilde{f}_{c, i_{max}} ∣ > 1, 25 \cdot Δ f$ .
The problem: the constants carry no units. Read in Hz, the damped step is capped at $5 \cdot 1 0^{- 5} Hz$ per iteration ( $2 \cdot 1 0^{- 3} Hz$ over all 40 iterations), so the tuning cannot move appreciably and the $1, 25 \cdot Δ f$ ( $\approx 0, 92 Hz$ ) failure check is unreachable; the whole clause would be inert. Read as normalized modulation rates $f / \tilde{r}_{s}$ (the variable in which the Formula (127) kernel frequencies are expressed), the same constants give a $0, 075 Hz$ damped per-iteration cap ( $\approx 2, 9 Hz$ over the 39 iterations), a $1, 5 \cdot 1 0^{- 4} Hz$ stop tolerance and a reachable failure check, all consistent with the clause’s purpose.
Evidence: dimensional analysis of the printed constants against the $0, 7324 Hz$ spectral resolution and the failure threshold.
Library behaviour: applies the constants as normalized modulation rates.
Status: unreported.

ECMA-418-2:2025 (4th edition), clause 9 introduction (broken cross-reference)

Location: clause 9, third paragraph of the introduction, on the HSA-based loudness prediction.
The print: “loudness scaling is improved by using HSA-based loudness prediction (see Clause 0)”.
The problem: “Clause 0” does not exist; the HSA-based loudness scaling is described in clause 9.1.10 (an unresolved field reference).
Evidence: the clause listing of the standard itself.
Library behaviour: none required (the intended target is unambiguous).
Status: unreported.

ISO/PAS 20065:2016, clause 5.3.4 (edge steepness of a distinct tone)

Location: clause 5.3.4, Formulae (10)/(11) (printed p. 9), the minimum edge steepness of a distinct tone.
The print: the two edges are scaled differently: $Δ L_{u} = (f_{T} /2) \cdot (L_{T max} - L_{u}) / (f_{T} - f_{u}) \geq 24 dB$ and $Δ L_{o} = f_{T} \cdot (L_{T max} - L_{o}) / (f_{o} - f_{T}) \geq 24 dB$ .
The problem: the parent standard DIN 45681:2005-03 prints $f_{T} / 2$ on both edges (Gleichungen (10)/(11), printed p. 14), and its executable Anhang J reference program does the same (Frequenz(i)/Sqr(2)). The two prints cannot both be satisfied. Neither ISO factor is the DIN one: on the lower edge $1/2 < 1/ 2$ , so the ISO print returns a level difference $2$ smaller and is therefore stricter; on the upper edge the divisor is absent altogether, so the ISO print returns $2$ larger and is more lenient. An earlier revision of this entry had the two directions the other way round and described the upper edge as “halved”, where in fact the divisor is missing rather than halved. Borderline tones with one-sided edge steepness between $24/ 2 = 17$ and $24 \cdot 2 = 34 dB/octave$ flip classification between the two readings.
Evidence: side-by-side comparison of the ISO print, the DIN 45681 print and the DIN Anhang J program. The DIN radicals are exactly the case the page rule exists for: pdftotext drops the √ glyph from both DIN formulae, so the extracted text reads f_T/2 and matches the ISO print, while the page itself reads f_T/√2. Verified on PDF page 13 (printed p. 9) of ISO/PAS 20065:2016 and PDF page 14 (printed p. 14) of DIN 45681:2005-03.
Library behaviour: follows the DIN/ $2$ reading (it matches the only executable reference), with the choice recorded in tone_audibility.py.
Status: unreported.

DIN 45681:2005-03, Anhang I, Tabelle I.6, row “6 FG”

Location: Anhang I, Beispiel I.2 (combustion engine, spectrum $j = 1$ ), Tabelle I.6, the combined row “6 FG” for the three tones $k = 6/7/8$ ( $592, 2$ / $629, 8$ / $643, 3 Hz$ , tone levels $78, 31$ / $75, 00$ / $79, 75 dB$ ).
The print: $L_{T} = 81, 11 dB$ together with $Δ L = 9, 12 dB$ (with $L_{S} = 59, 53$ , $L_{G} = 76, 16$ , $a_{v} = - 2, 40$ at $592, 2 Hz$ ).
The problem: the two cells contradict each other. The printed $Δ L = 9, 12 dB$ only reproduces from the plain Formula (17) energy sum of the three tone levels ( $82, 8734 dB$ ): $82, 87 - 76, 16 + 2, 40 = 9, 11$ . The printed $L_{T} = 81, 11 dB$ is that same sum less exactly $1, 763 dB$ , and taken at face value it would give $Δ L = 7, 35 dB$ .
Evidence: recomputation from the printed per-tone levels of Tabelle I.6. The offset is the discriminator and it is a constant, not a deduplication: $82, 8734 - 81, 11 = 1, 763 dB$ , and $1, 76 dB$ is $10 lo g_{10} 1, 5$ , the standard’s own Hanning effective-bandwidth correction (clause 5.3.2). The same offset appears in the “5 FG” row of Tabelle I.10 (printed p. 46), where the two member tones at $705, 2$ and $732, 1 Hz$ have $L_{T} = 55, 12$ and $54, 23 dB$ , sum to $57, 708 dB$ , and are printed as $55, 95 dB$ , $1, 758 dB$ lower, and there the printed $Δ L = 3, 22 dB$ follows the printed $L_{T}$ exactly ( $55, 95 - 55, 28 + 2, 55 = 3, 22$ ), so the Tabelle I.10 row is internally consistent and the Tabelle I.6 row is not. The third combined row, “2 FG” of the same Tabelle I.6, carries no offset at all: its three member levels $64, 56$ / $67, 96$ / $68, 63 dB$ sum to $72, 149 dB$ against a printed $72, 15 dB$ , and its $Δ L$ follows. A previous revision of this entry attributed the $81, 11 dB$ cell to the Anmerkung 2 shared-line deduplication; that diagnosis is unsupported, because a deduplication removes an arbitrary amount of energy while all the offsets observed here are the same 1,76 dB. Verified on PDF page 41 (printed p. 41) and PDF page 46 (printed p. 46) of DIN 45681:2005-03.
Library behaviour: combined_tone_level follows Anmerkung 2 (shared lines counted once), which reproduces the printed “2 FG” oracle; for the “6 FG” row only the $Δ L$ chain is pinned, with the contradiction recorded in tests/reference_data/.
Status: unreported.

DIN 45681:2005-03, Anhang I, Tabellen I.2 and I.10 (wrong spectrum index in a column header)

Location: Anhang I, the column headers of Tabelle I.2 (printed p. 37, spectrum $j = 2$ ) and Tabelle I.10 (printed p. 46, spectrum $j = 24$ ).
The print: every column of Tabelle I.2 is subscripted with the spectrum index 2 (f_T 2,k, f_1 2,k, f_2 2,k, L_S 2,k, L_T 2,k, L_G 2,k, a_v 2,k, u_2,k) except the audibility column, which is headed ΔL_1,k. Every column of Tabelle I.10 is subscripted 24 (f_T 24,k, ΔL 24,k, f_1 24,k, f_2 24,k, L_S 24,k, L_T 24,k, L_G 24,k, u 24,k) except the masking column, which is headed a_v 1,k.
The problem: both tables carry the spectrum index of the first spectrum in one column. Tabelle I.2’s own caption reads “des zweiten Spektrums (j = 2)” and Tabelle I.10’s “des 24. Spektrums (j = 24)”, and the body values belong to those spectra: the $Δ L$ column of Tabelle I.2 is the audibility of the $j = 2$ tones ( $8, 53 dB$ at $627, 2 Hz$ , which the Anmerkung below the table calls “die maßgebliche Differenz ΔL_2”), and the $a_{v}$ column of Tabelle I.10 is the masking index of the $j = 24$ tones. The index 1 is right in exactly one table of the annex, Tabelle I.6, which is the $j = 1$ table of Beispiel I.2 and carries both ΔL_1,k and a_v 1,k legitimately.
Evidence: the tables’ own captions, their neighbouring column subscripts, and the Anmerkung under each. Verified on PDF page 37 (printed p. 37), PDF page 46 (printed p. 46), and PDF page 41 (printed p. 41) of DIN 45681:2005-03.
Library behaviour: none needed; the numbers are unaffected. The regression fixtures index both tables by their caption’s spectrum.
Status: unreported.

IEC 60268-3:2013, clause 14.12.9.2 f) (DIM denominator)

Location: clause 14.12.9.2, item f) (printed p. 39), the formula for the dynamic intermodulation distortion $d_{DIM}$ .
The print: $d_{DIM} = (\sum_{i = 1}^{9} U_{i}^{'}^{2})^{1/2} / U_{2} \times 100 %$ .
The problem: the denominator is one of the nine terms of its own numerator. Table 2 of the same clause (printed p. 38) defines $U_{2}$ as the intermodulation component at $f_{s} - 2 f_{q} = 8, 70 kHz$ , and item d) defines $U_{1}, U_{2}, \dots U_{i}$ as exactly those components, so the sum $i = 1 \dots 9$ runs over $U_{1} \dots U_{9}$ and includes $U_{2}$ . Meanwhile the defining clause 14.12.9.1 states the ratio of the r.m.s. sum of the Table 2 intermodulation product voltages “to the amplitude of the output voltage at the frequency f_s”, i.e. the 15 kHz sine component $U_{s}$ , the Otala convention, and item d) measures “the amplitudes of the sinusoidal signal $U_{s}$ ” precisely so that it can be used, which the f) formula then never does. The denominator should be $U_{s}$ . An earlier revision of this entry said that “U2 is used throughout 14.12 for the total output voltage”; that is false, in both the English and the French print.
Evidence: Table 2, item d) and item f) read together in both language columns of the bilingual edition; the historical DIM literature (Otala) defines the ratio to the sine amplitude. Verified on PDF page 41 (printed p. 39), PDF page 40 (printed p. 38), which carries Table 2, and PDF page 102 (printed p. 100), which carries the same item f) in the French column, of IEC 60268-3:2013.
Library behaviour: follows the 14.12.9.1 definition (reference = the output amplitude at $f_{s}$ ), with a code comment at the reference measurement in distortion.py.
Status: unreported.

UNE-EN 61043:1999, clause 6.1 (class 2 frequency range dropped in translation)

Location: clause 6.1 “Rango de frecuencias”, the class 2 sentence, of UNE-EN 61043 (April 1999), which declares itself “la versión oficial, en español, de la Norma Europea EN 61043 de enero 1994, que a su vez adopta la Norma Internacional CEI 61043:1993”.
The print: a single sentence, “Los procesadores de clase 2 deberán cubrir, al menos, el rango desde 45 Hz a 5,6 kHz en bandas de octava.”
The problem: the EN/IEC text gives class 2 processors two alternative ranges, not one: “Class 2 processors shall, at least, cover the range from 45 Hz to 7,1 kHz in one-third octave bands, or the range from 45 Hz to 5,6 kHz in one octave bands” (BS EN 61043:1994, clause 6.1). The translation drops the first alternative. The omission is normative rather than editorial: it removes one of the two ways clause 6.1 can be satisfied, and a reader of the Spanish text alone would conclude that class 2 is defined over octave bands, so that a one-third-octave chain verified over the 22 tabulated bands from 50 Hz to $6, 3 kHz$ could not attest class 2 over its full range.
Evidence: side-by-side reading of clause 6.1 in both prints. The class 1 sentence is word-for-word equivalent in the two documents, so the divergence is confined to the class 2 sentence. The Spanish print also contradicts itself: its Table 2 tabulates the pressure-residual intensity index for class 2 processors at all 22 one-third-octave centres, and its faithfully translated Note 2 (“Para procesadores con análisis en bandas de octavas únicamente, los requisitos se aplican únicamente a las frecuencias centrales de las bandas de octava”) carves out octave-only processors as a special case. Both are redundant if every class 2 processor is an octave-band one.
Library behaviour: implements the EN/IEC reading. verify_intensity_class treats the full 22-band one-third-octave set as attesting either class, and the 7-band octave set (63 Hz to 4 kHz) as a class 2 alternative that never attests class 1, with both branches pinned by regression tests (tests/emission/test_intensity_compliance.py).
Status: unreported (national translation, not the issuing body’s text).

ISO/PAS 1996-3:2022, Clause 5 (cross-references of r and d)

Location: Clause 5, Formula (2), the definitions of the symbols of the prominence $P = 3 lo g_{10} [r / (dB/s)] + 2 lo g_{10} (d / dB)$ .
The print: “r is the onset rate (OR) as defined in 3.4” and “d is the level difference (LD) as defined in 3.5”.
The problem: the two cross-references are swapped. The document’s own terms and definitions set 3.4 as the level difference LD (“difference in decibels of L_pAF between the level of the end point L_e and the level of the starting point L_s of the onset”) and 3.5 as the onset rate OR (“slope in decibels per second of the straight line that gives the best approximation to the onset”). Read literally, Formula (2) would take three times the logarithm of a level difference plus twice the logarithm of a slope, inverting the weights the method assigns to the two quantities. The spelled-out names in the same list (“the onset rate (OR)”, “the level difference (LD)”) and the units given for each (“dB/s” for $r$ , “dB” for $d$ ) make the intended reading unambiguous.
Evidence: side-by-side reading of 3.4, 3.5 and the Clause 5 symbol list; the units printed with each symbol contradict the clause numbers printed with them.
Library behaviour: implements the spelled-out reading, weighting the onset rate by 3 and the level difference by 2 (predicted_prominence in impulsive_sound.py), which is also the NT ACOU 112:2002 form the PAS carries over.
Status: unreported.

ISO 9613-2:1996, Table 2 (15 °C / 80 % / 1 kHz cell)

Location: Table 2, “Atmospheric attenuation coefficient α for octave bands of noise”, row 15 °C / 80 % relative humidity, column 1 kHz.
The print: $α = 4, 1 dB/km$ .
The problem: Table 2 is a rounded extract of ISO 9613-1, to which the clause itself defers (“For values of α at atmospheric conditions not covered in table 2, see ISO 9613-1”). Evaluating the ISO 9613-1 pure-tone formula at 1 kHz, $15^{\circ} C$ , $80 %$ RH and $101, 325 kPa$ gives $4, 1511 dB/km$ , which rounds to $4, 2$ , not the printed $4, 1$ . The neighbouring cells of the same row round correctly (2 kHz: $8, 338$ -> printed $8, 3$ ; 4 kHz: $23, 86$ -> $23, 7$ at the exact band centre), as do the 1 kHz cells of the other rows ( $15^{\circ} C$ / $50 %$ : $4, 164$ -> printed $4, 2$ ), so the defect is confined to this cell.
Evidence: independent evaluation of the ISO 9613-1 coefficient at both the nominal and the exact band-centre frequency ( $4, 1511 dB/km$ either way, 1 kHz being both).
Library behaviour: unaffected. The library never reads Table 2: it computes $A_{atm}$ from the ISO 9613-1 formula directly (air_absorption.py), so it yields $4, 15 dB/km$ for this condition.
Status: unreported.

ANSI S3.5-1997, Annex C worked examples (official WG S3-79 errata)

Not verified against the page. ANSI S3.5-1997 is not held locally (the R package SII vignette is held, not the standard), so what this entry calls “the print” is the working group’s own description of it, not a page this project has read. The recomputations below are independent and do reproduce, but the printed characters rest on the errata list alone. The standard is on the maintainer’s pending-acquisition list; when a copy arrives the entry is to be re-verified against the print of printed pp. 21-22 and this notice removed.

Location: Annex C, Table C.1 (octave-band worked example, p. 21) and Table C.2 (one-third-octave worked example, p. 22) of the 1997 printing.
The print (per the working group’s errata): (a) Table C.1, row $i = 5$ , the level-distortion factor $L_{i}$ under Step 6 is printed as $0.10$ ; (b) Table C.2, first row, the self-speech-masking slope $C_{i}$ is printed as $- 45.59$ .
The problem: both cells contradict the standard’s own normative formulas. (a) Clause 5.7 with the example’s inputs ( $E_{5}^{'} = 20 dB$ , $U_{5} = 9.33 dB$ ) gives $L_{5} = 1 - (20 - 9.33 - 10) /160 = 0.9958$ , which prints to two decimals as $1.00$ , not $0.10$ . (b) Clause 5.4 with the example’s inputs ( $B_{1} = 40 dB$ , $f_{1} = 160 Hz$ ) gives $C_{1} = - 80 + 0.6 (40 + 10 lo g_{10} 160 - 6.353) = - 46.587$ , which prints as $- 46.59$ , not $- 45.59$ ; the example’s $Z_{i}$ column is only consistent with the corrected slope ( $Z_{2}$ recomputes to $34.658$ = printed 34.66 dB, whereas the misprinted slope would give 34.76 dB). The Table C.1 example is the octave-band procedure and the Table C.2 example the one-third-octave procedure, so one cell of each is affected.
Evidence: the official errata list published by ASA Working Group S3-79, the committee that maintains ANSI S3.5, on its support site (sii.to): “Page 21, Table C1, row i=5, column Li under Step 6: the value printed as 0.10 should be changed to 1.00” and “Page 22, Table C2, the first row of numbers, value −45.59 should be −46.59”; plus independent recomputation of both cells from the normative clauses (above). The same list carries five further corrections (a reference spelling, the Tables 1-4 caption wording recorded in the next entry, the insertion gain $G_{i}$ missing from Eq. 23, and two Annex B fixes, a cross-reference “B16” that should read “B15” and a wording change about the audio-visual approximation); none of those touches a formula this library implements. The source is the WG S3-79 errata list at sii.to/html/errata.html (captured 2026-07-30, re-checked live 2026-08-04). It is not the printed page and cannot substitute for it, which is why this entry carries the notice above.
Library behaviour: unaffected; the library computes the corrected values from the normative clauses and always did. Its Annex C.2 anchors (tests/reference_data/, ANSIS3_5_ANNEX_C1* and ANSIS3_5_ANNEX_C2*) pin the errata-consistent chain of both examples, cross-checked to double precision against the working group’s own reference implementation SII.C and its published test-case results. The Table C.1 cell is pinned directly: the level-distortion factor of clause 5.7 for row $i = 5$ of the Annex C.1 octave-band example computes to $0.99581$ , which prints as the corrected $1.00$ .
Status: published corrections by the issuing working group; nothing to report upstream.

ANSI S3.5-1997, captions of Tables 1 to 4 (official WG S3-79 erratum)

Not verified against the page. As with the entry above, ANSI S3.5-1997 is not held locally, so the wording of the four captions is taken from the working group’s errata list rather than from a page this project has read. The argument that the tables carry no threshold column is independent and does hold against the transcribed constants. Re-verify against the print of printed pp. 3-5 when the standard is acquired.

Location: the captions of Tables 1, 2, 3 and 4 (pp. 3-5 of the 1997 printing), the constant tables of the four band procedures: critical band (21 bands), equally-contributing critical band (17 bands), one-third octave (18 bands) and octave (6 bands).
The print (per the working group’s errata): each caption lists the quantities the table tabulates and includes the phrase “hearing threshold levels,”.
The problem: none of the four tables tabulates a hearing threshold level. Each carries the band centre frequency (and, for Tables 1, 2 and 4, the band limits), the band-importance function $I_{i}$ , the standard speech spectrum level $U_{i}$ by vocal effort and the reference internal noise spectrum level $X_{i}$ . The hearing threshold level $T_{i}^{'}$ is a user input to the procedure (clause 5.5, where the equivalent internal noise spectrum level is $X_{i}^{'} = X_{i} + T_{i}^{'}$ ), which is exactly the quantity the caption invites the reader to look for in the table and to confuse with $X_{i}$ .
Evidence: the official errata list published by ASA Working Group S3-79, the committee that maintains ANSI S3.5, on its support site (sii.to): “Pages 3-5, Tables 1-4: In each of the figure captions the phrase ‘hearing threshold levels,’ should be deleted” (the WG S3-79 errata list at sii.to/html/errata.html, captured 2026-07-30, re-checked live 2026-08-04; an earlier revision of this entry dropped the word “figure” from the quotation); plus the tables themselves, which have no such column.
Library behaviour: unaffected. The four tables are implemented with the columns they actually carry, exposed per procedure by sii_procedure() as band_importance, speech_spectrum ( $U_{i}$ ) and internal_noise ( $X_{i}$ ), and the hearing threshold stays the threshold= argument of speech_intelligibility_index (src/phonometry/speech/sii.py).
Status: published correction by the issuing working group; nothing to report upstream.

NORAH2 rotorcraft guidance SC01.D1.5d (EASA.2020.FC.06), Eq. (27)

Location: section A.4.2, Eq. (27) (atmospheric absorption coefficient) and the sentence defining its symbols, printed p. 21.
The print: Eq. (27) pairs the coefficient $6.6928 \cdot 1 0^{- 6}$ with $f_{r O}$ and $1.3415 \cdot 1 0^{- 6}$ with $f_{r N}$ , and the sentence below reads “the variables f_rN = 75692 Hz and f_rO = 630.7 Hz represent the vibrational relaxation frequencies of oxygen and nitrogen respectively”.
The problem: the two subscripts are swapped in the definition sentence. The values match the names it gives them (75 692 Hz is the oxygen relaxation frequency and 630.7 Hz the nitrogen one at the reference conditions), but they are assigned to the opposite symbols, so the equation as printed multiplies the oxygen coefficient by the nitrogen relaxation frequency and vice versa. Evaluated that way it gives 14.2 dB/km at 500 Hz against the guidance’s own Table 4 value of 3.1 dB/km; with $f_{r O}$ and $f_{r N}$ exchanged it gives 3.07 dB/km, reproducing Table 4 and the ISO 9613-1 pure-tone coefficient to 0.02 dB/km. An earlier revision of this entry quoted the printed value as 14.3 dB/km and framed the defect as a wrong pairing of the coefficients rather than as swapped subscripts in the definition.
Evidence: numeric evaluation of Eq. (27) with the printed assignment and with the assignment exchanged, against the Table 4 500 Hz cell on the same page. Verified on PDF page 20 (printed p. 21) of NORAH2 SC01.D1.5d (EASA.2020.FC.06):2024.
Library behaviour: implements the correct pairing; the module docstring carries a defensive note so the misprint is not transcribed as a “fix”.
Status: unreported.

NORAH2 rotorcraft guidance SC01.D1.5d (EASA.2020.FC.06), Eq. (21)

Location: section A.3.3, Eq. (21) (flight path angle).
The print: $γ = acos (Δ Z /Δ S)$ .
The problem: the arccosine of the climb-to-path ratio returns the complement of the path angle ( $9 0^{\circ}$ in level flight, where $γ$ must be $0^{\circ}$ ) and contradicts the guidance’s own use of $γ$ as the climb/descent angle throughout section A.3. ECAC Doc 32, 1st ed., Eq. (10) prints the correct form, $γ = atan (Δ Z /Δ S)$ with the horizontal $Δ S$ of its Eq. (8).
Evidence: evaluation in level flight; cross-check against Doc 32 Eq. (10) and against the NORAH2 prototype input files, whose Vang columns are climb/descent angles ( $0^{\circ}$ in level segments).
Library behaviour: flight_path_kinematics implements the Doc 32 atan form; the result docstring carries the defensive note.
Status: unreported.

NORAH2 rotorcraft guidance SC01.D1.5d (EASA.2020.FC.06), §A.3.1 triangulation

Location: section A.3.1, steps 2 to 4 (flight-condition interpolation), against the triangulation lookup tables shipped with the NORAH2 database (*_triangulation.int).
The print: steps 2 and 3 normalise the database conditions (spans, with $F_{f c} = 2$ on the path angle) and step 4 computes “the Delaunay triangulation for the database flight conditions γ̄_j and V̄_j”, i.e. of the normalised points, offering a lookup table as an equivalent.
The problem: the lookup tables shipped with the database (which the guidance says are part of the hemisphere data and should not be edited) are the Delaunay triangulation of the raw $(V, γ)$ conditions, not of the normalised ones: for the R22 set, 14 of the 27 shipped triangles differ from the Delaunay triangulation of the normalised conditions. A Delaunay triangulation is not invariant under the anisotropic normalisation, so the two prescriptions select different enveloping triangles for part of the envelope. The distance weights of Eq. (7)/(8) do use the normalised coordinates in the prototype (verified against its blended outputs).
Evidence: recomputation of both triangulations for the R22 database; bin-for-bin reproduction of the prototype’s per-step hemisphere selection with the shipped tables, and of its blended levels with normalised-space weights, to 0.05 dB.
Library behaviour: flight_condition_weights follows the printed method (Delaunay of the normalised conditions) by default and accepts the database lookup table via triangles, which reproduces the reference implementation exactly.
Status: unreported.

NORAH2 rotorcraft guidance SC01.D1.5d (EASA.2020.FC.06), Eq. (46)

Location: section A.4.5, Eq. (46) (source-side ground effect weighted by diffraction).
The print: the weighting exponent reads $(Δ L_{g, s^{'}} - Δ L_{d, s}) /20$ .
The problem: no term $Δ L_{g, s^{'}}$ exists; the prose directly below the equation defines $Δ L_{d, s^{'}}$ as “the attenuation due to the diffraction between the image source S′ and R”, the receiver-side companion Eq. (47) prints the parallel term correctly as $Δ L_{d, r^{'}}$ , and the CNOSSOS-EU method the section is based on writes $Δ_{ground} (S, O)$ with $Δ_{dif} (S^{'}, R)$ in that position. The subscript $g$ is a misprint for $d$ .
Evidence: internal consistency of the section (its own prose and Eq. (47)) and the CNOSSOS-EU source of the equations.
Library behaviour: implements the image-source diffraction term $Δ L_{d, s^{'}}$ as defined by the prose.
Status: unreported.

NORAH2 rotorcraft guidance SC01.D1.5d (EASA.2020.FC.06), §A.4.5 cross-references

Location: section A.4.5, the definitions under Eq. (46) (printed p. 32) and Eq. (47) (printed p. 33).
The print: four cross-references to eq. 44, in three different wordings: “calculated as per eq. 44” for $Δ L_{d, s^{'}}$ and again for $Δ L_{d, s}$ under Eq. (46); “calculated as in eq. 44” for $Δ L_{d, r^{'}}$ under Eq. (47); and “calculated as in Subsection eq. 44” for $Δ L_{d, s}$ under Eq. (47). An earlier revision of this entry quoted all four with the first wording.
The problem: Eq. (44) is the multiple-diffraction coefficient $C^{''}$ ; the attenuation due to diffraction is Eq. (42). All four cross-references point at the auxiliary coefficient instead of the formula they describe, and the fourth also carries a dangling “Subsection” with no subsection number after it.
Evidence: the terms are attenuations in dB, which only Eq. (42) produces; Eq. (44) is a dimensionless coefficient consumed by Eq. (42). Verified on PDF pages 31 and 32 (printed pp. 32 and 33) of NORAH2 SC01.D1.5d (EASA.2020.FC.06):2024.
Library behaviour: evaluates the image-path and direct diffraction terms with Eq. (42), using Eq. (44) for $C^{''}$ inside it.
Status: unreported.

NORAH2 rotorcraft guidance SC01.D1.5d (EASA.2020.FC.06), §A.3.5 Approach 3 (full-rpm idle base)

Location: section A.3.5, Approach 3, step 3 (printed p. 18), against the “Fl. idle” row of Table 3 (printed pp. 18-19).
The print: the step reads “add offset of 12 dB* to derive out of ground hover from the in-ground hover disk, -12 dB* to derive reduced-rpm idle from in-ground hover disk, and -2.5 dB* to derive full-rpm idle from out of ground hover”; the table prints $L A_{FL.idle} (θ) = L A_{HIGE} (θ) - 2.5 dB^{*}$ .
The problem: the prose derives full-rpm idle from out-of-ground hover where the table derives it from in-ground hover, and the two prescriptions land 12 dB apart (via the prose, $L A_{HOGE} (θ) - 2.5 = L A_{HIGE} (θ) + 9.5$ ; via the table, $L A_{HIGE} (θ) - 2.5$ ). Only the table keeps the physical ordering of the conditions (full-rpm idle above reduced-rpm idle, both below in-ground hover). The paragraph that introduces these phases, at the end of section A.3.3 (printed p. 17), is itself left unfinished (“For specific phases of a flight such as, turns, hover, taxiing”), pointing at an editing pass the section did not get.
Evidence: the corrections shipped with the V2.0.74 public database are all relative to the in-ground-hover disk (Fullrpmidle -2 in every type’s interpolation lookup file), agreeing with the table and not with the prose. Verified on PDF pages 16, 17 and 18 (printed pp. 17, 18 and 19) of NORAH2 SC01.D1.5d (EASA.2020.FC.06):2024.
Library behaviour: hover_derived_hemisphere applies every Table 3 offset from the in-ground-hover hemisphere, as the table prints; the docstring states the base condition explicitly.
Status: unreported.

NORAH2 rotorcraft guidance SC01.D1.5d (EASA.2020.FC.06), §A.3.5 taxi assignment

Location: section A.3.5, last paragraph (printed p. 19).
The print: “To include taxiing for helicopters with and without wheels into the noise calculation the measured and derived hemispheres for in-ground hover and full-rpm idle respectively should be employed.”
The problem: read literally, the “respectively” pairs the wheeled helicopter with the in-ground-hover source and the wheel-less one with full-rpm idle, which is the reverse of the operations it models: a helicopter without wheels can only taxi by hovering in ground effect, and a wheeled helicopter ground-taxis on its wheels with the rotor at governed idle, not producing lift. The two lists read as transposed. No oracle settles it (the public release ships no taxi verification case), so the pairing is corrected from the physics of the operations alone.
Evidence: internal comparison of the two prose lists against the operations they name. Verified on PDF page 18 (printed p. 19) of NORAH2 SC01.D1.5d (EASA.2020.FC.06):2024.
Library behaviour: no function is affected (the rule selects between two hemispheres the reader has already built); the rotorcraft guide documents the physical pairing, wheel-less taxi on the in-ground-hover hemisphere and wheeled taxi on the full-rpm-idle one, with this caveat.
Status: unreported.

NORAH2 rotorcraft guidance SC01.D1.5d (EASA.2020.FC.06), Table 3 offsets vs the shipped corrections

Location: Table 3, Approach 3 column (printed pp. 18-19), against the &CORRECTIONS block of the interpolation lookup files shipped with the NORAH2 V2.0.74 public release.
The print: offsets of +12 dB* (out-of-ground hover), -12 dB* (reduced-rpm idle) and -2.5 dB* (full-rpm idle) from the in-ground-hover disk, with the asterisked note that they were derived from measurements with inverted microphones on ground plates and “may not be valid for other microphone setups”.
The problem: the reference database the guidance builds on ships different values: every one of the eleven per-type triangulation lookup files (*_triangulation.int) of the public release carries Corr_dB 8, -10 and -2 for the same three operations, so the published constants and the database disagree by 4, 2 and 0.5 dB. The guidance, whose section A.3.1 declares the shipped lookup data part of the hemisphere database and not to be edited, does not mention the difference, and its note questions the validity of the published values without naming the ones actually shipped.
Evidence: the identical &CORRECTIONS blocks of the eleven triangulation lookup files (*_triangulation.int) of the V2.0.74 public release; the published constants verified on PDF pages 17 and 18 (printed pp. 18 and 19) of NORAH2 SC01.D1.5d (EASA.2020.FC.06):2024.
Library behaviour: hover_derived_hemisphere defaults to the published Table 3 constants and accepts a measured or database correction as offset_db; the end-to-end hover verification case passes the database’s +8 dB explicitly, and the docstring records the divergence.
Status: unreported.

RANDI 3.1 Physics Description (NRL, Breeding et al.), Table 2

Location: Table 2 (representative ship source levels).
The print: two cells deviate from the report’s own Eqs. (2) to (5) evaluated with the Table 1 average lengths and speeds: the Merchant value at 25 Hz (about 3 dB high) and the Tanker value at 300 Hz (about 1 dB low). The Fishing Vessel row is not reproducible from the Table 1 averages at all (a constant offset of about 3.8 dB suggests different assumed inputs).
The problem: the report does not state the exact inputs used for Table 2, and two cells contradict its own equations while every Large Tanker and Super Tanker cell agrees to 0.06 dB.
Evidence: recomputation of all 25 cells from Eqs. (2) to (5).
Library behaviour: the regression test pins the reproducible rows and excludes the contradicting cells with the rationale in the test.
Status: unreported (technical report rather than a standard).

Osses, García & Kohlrausch (2016), fluctuation-strength model, Eq. (3)

Location: Eq. (3), the critical-band-rate (Bark) transformation of the excitation-pattern front-end.
The print: $z (f) = 13 \cdot arctan (0.76 \cdot 1 0^{- 4} \cdot f) + 3.5 \cdot arctan ((f /7500)^{2})$ .
The problem: the first coefficient is the Zwicker-Terhardt $0.76 \cdot 1 0^{- 3}$ with the exponent misprinted. The paper’s own anchors disprove the print: it states $0.5 Bark = 50 Hz$ and $23.5 Bark = 13.2 kHz$ (section 2.1.2) and $15 Bark = 2.7 kHz$ (section 3.1), all of which require $1 0^{- 3}$ . With $1 0^{- 4}$ , $z (1 kHz) = 1.05$ instead of $8.51 Bark$ and the model’s 47 filter centres would span 491 Hz to 20 kHz instead of 50 Hz to 13.2 kHz.
Evidence: evaluation of Eq. (3) under both exponents against the paper’s printed Bark/frequency anchors. The printed section 2.1.2 range “0.5 Bark (50 Hz) to 23.5 Bark (13.2 kHz)” and the section 3.1 anchor “15 Bark (2.7 kHz)” all reproduce under the Zwicker-Terhardt $0.76 \cdot 1 0^{- 3}$ (50.6 Hz, 13.07 kHz and 2.71 kHz) and none of them under the printed exponent. Verified on PDF page 4 (printed p. 4) of Osses, García & Kohlrausch, ICA:2016, with the anchors on PDF page 7 (printed p. 7) of the same paper.
Library behaviour: implements $0.76 \cdot 1 0^{- 3}$ with a note at the formula; the carrier-frequency sweep test would catch a regression to the printed value (fluctuation_strength.py).
Status: unreported (conference paper rather than a standard).

Medwin & Clay, Fundamentals of Acoustical Oceanography (1998), Eq. (3.4.30) (boric-acid coefficient)

Location: the Francois-Garrison boric-acid term as transcribed by the textbook, Eq. (3.4.30), printed p. 110. An earlier revision of this entry cited Eq. 3.4.29, which is the total-absorption sum of the three terms on printed p. 109; the boric-acid block is the equation after it.
The print: $A_{1} = (8.68/ c) \cdot 1 0^{0.78 pH - 5} dB km^{- 1} kHz^{- 1}$ .
The problem: the original paper (Francois & Garrison 1982, JASA 72, Part II, Eq. (10) and Fig. 7) prints 8.86; the digits are transposed. Only 8.86 reproduces the paper’s own Table IV: with 8.68 the boric-dominated cells at $0.6$ to 30 kHz sit up to $1.7 %$ below the printed totals (worst relative case 2 kHz, 10 °C, $S = 35$ : $0.1209$ vs the printed 0.123 dB/km).
Evidence: recomputation of all sampled Table IV cells under both coefficients against the paper’s printed values. Verified on PDF page 131 (printed p. 110) of Medwin & Clay, Fundamentals of Acoustical Oceanography (1998), and on PDF pages 8 and 9 (printed pp. 1886 and 1887) of Francois & Garrison (1982), JASA 72, Part II, which print the paper’s own $A_{1} = (8.86/ c) \cdot 1 0^{0.78 pH - 5}$ .
Library behaviour: implements the paper’s 8.86 with a defensive note; the pinned Table IV set includes the boric-dominated rows.
Status: unreported (textbook rather than a standard).

Medwin & Clay (1998), Eq. (3.4.30) (sound speed printed as q)

Location: the same Eq. (3.4.30) block, printed p. 110, its last line.
The print: $q = 1412 + 3.21 T + 1.19 S + 0.0167 z m/s$ .
The problem: the quantity the block needs is the sound speed $c$ , which is what the two lines above it divide by ( $A_{1} = 8.68/ c$ , and $A_{2} = 21.44 S / c$ in the magnesium-sulfate block on the same page). No symbol $q$ is defined anywhere in the section, so the transcribed system is not closed: a reader following the printed symbols has no value for $c$ . Francois & Garrison 1982 Part II prints the same polynomial as $c = 1412 + 3.21 T + 1.19 S + 0.0167 D$ , introduced by “where c is the sound speed (m/s), given approximately by”.
Evidence: the block’s own use of $c$ two lines above, and the source paper. Verified on PDF page 131 (printed p. 110) and PDF page 130 (printed p. 109) of Medwin & Clay (1998), and of PDF page 8 (printed p. 1886) of Francois & Garrison 1982 Part II (JASA 72).
Library behaviour: unaffected; the absorption model takes the sound speed from the same polynomial under the name c.
Status: unreported (textbook rather than a standard).

Maa (1998), “Potential of microperforated panel absorber”, JASA 104(5), Eq. (5b)

Location: Eq. (5b), the mass-reactance coefficient of the microperforated panel, printed as $k_{m} = 1 + [1 + k^{2} /2]^{- 1/2} + 0.85 d / t$ .
The print: the first bracket term reads $(1 + k^{2} /2)^{- 1/2}$ .
The problem: the same paper’s Eq. (4), from which (5b) is factored, prints the term as $(3^{2} + k^{2} /2)^{- 1/2}$ , and only that form reproduces the Crandall low- $k$ limit $Z_{1} \to (4/3) j ω ρ_{0} t$ of the paper’s own Eq. (3a): at $k \to 0$ the printed (5b) gives an internal mass factor of 2 instead of 4/3. The paper’s own Fig. 1 confirms it: with $0.85 \cdot d / t = 0.85$ the plotted $k_{m}$ starts near $2.2$ ( $= 4/3 + 0.85$ ) at $k = 0.1$ , not at $2.85$ .
Evidence: recomputation of both bracket variants against Eq. (4), Eq. (3a) and the Fig. 1 curve; the exact Bessel solution of Eq. (2) agrees with Eq. (4) within Maa’s stated $\sim 6 %$ only with the $3^{2}$ form (the 1 form errs by $> 30 %$ at low $k$ ). Verified on PDF page 2 (printed p. 2862) of Maa (1998), “Potential of microperforated panel absorber”, JASA 104(5), which carries Eq. (4) and Eq. (5b) fifteen lines apart on the same column.
Library behaviour: implements the exact Eq. (2) (no approximation), so the misprint does not enter the code; the regression test test_maa_exact_vs_wide_range_approximation pins the exact solution to the corrected Eq. (4) form.
Status: unreported (journal paper; the correct form appears in Maa’s earlier 1975/1987 papers and in secondary literature).

Jiménez, Groby, Pagneux & Romero-García (2017), Appl. Sci. 7(6), 618, Eqs. (7)-(8)

Location: Eqs. (7) and (8), the rectangular-duct visco-thermal effective density and bulk modulus (Stinson’s series, used for the square necks and cavities of the slit + Helmholtz-resonator absorber).
The print: the leading normalising constant of both series is 4: $ρ_{eff} = - ρ_{0} \cdot a^{2} b^{2} / (4 \cdot G_{ρ}^{2} \cdot Σ)$ and the matching $4 \cdot (γ - 1) \cdot G_{κ}^{2} / (a^{2} b^{2})$ factor inside $κ_{eff}$ .
The problem: the correct constant is 64 (a factor-16 error). Only 64 reproduces the exact limits of the model: as the boundary layers vanish $ρ_{eff} \to ρ_{0}$ and $κ_{eff} \to κ_{0}$ (the printed 4 gives $16 \cdot ρ_{0}$ ), and at DC the square duct’s $j ω \cdot ρ_{eff}$ tends to the exact Shah-London Poiseuille flow resistivity: the series value $a^{6} / (64 \cdot S_{0}) = 28.4542$ matches $f R e /2 = 28.455$ (in units of $η / a^{2}$ ), where $S_{0}$ is the double transverse-mode sum at $G = 0$ ; the printed 4 gives sixteen times that.
Evidence: evaluation of both constants against the boundary-layer-free limits and the Shah-London exact square-duct value; the wide-duct limit of the series also only matches the papers’ own slit model (Eq. (6)) with 64.
Library behaviour: implements 64 with a docstring note; the limits are pinned in tests/materials/absorbers/test_slow_sound.py and the conformance check “Poiseuille limit (Stinson 1991)”.
Status: unreported (journal paper rather than a standard).

Jiménez et al. (2017), Appl. Sci. 7(6), 618 / Sci. Rep. 7, 5389, slit-radiation term

Location: Appl. Sci. Eq. (3), the characteristic radiation impedance of the slits, and the identical Methods reprint in the metadiffusers paper (Sci. Rep. 7, 5389, Eq. (5)).
The print: $Z_{Δ l_{slit}} = - iω \cdot Δ l_{slit} \cdot ρ_{0} / (ϕt \cdot S_{0})$ .
The problem: the term models the added radiation mass of the slit mouth, but the printed $- iω$ prefactor is an opposite-time-convention ( $e^{- iω t}$ ) expression inconsistent with the papers’ otherwise $e^{+ iω t}$ transfer-matrix chain (the $+ i$ off-diagonal slit matrices of Appl. Sci. Eq. (2) and the $- i$ cotangent-type resonator impedance). Transcribed literally into that chain, the correction raises the slit-panel resonance where an added mass must lower it: for a 1 mm slit with a 30 mm lattice step and 50 mm period the absorption peak moves from 378.6 Hz to 386.8 Hz as printed, against 370.8 Hz with the mass sign. The neck end corrections of the same model behave correctly (they lower the resonator resonance).
Evidence: numerical evaluation of both signs of the correction against the uncorrected panel; the direction of the neck end corrections of the same papers as the consistent control.
Library behaviour: uses the added-mass sign ( $+ j ω$ in the $e^{+ j ω t}$ convention of the library), conjugating the printed term exactly as it conjugates the papers’ Stinson duct series; direction and peak are pinned by test_slit_radiation_correction_lowers_resonance in tests/materials/absorbers/test_slow_sound.py.
Status: unreported (journal papers rather than standards).

Attenborough & Van Renterghem, Predicting Outdoor Sound 2e (2021), Table 5.1

Location: Table 5.1, “Coefficient and exponent values in the Delany and Bazley, Miki and modified Miki models”, row “Miki [6,7]”, coefficient $r$ .
The print: $r = 0.0109$ .
The problem: the original source (Miki 1990, J. Acoust. Soc. Jpn (E) 11(1), Eq. (34)) prints $β (f) = (ω / c_{0}) [1 + 0.109 \cdot (f / σ)^{- 0.618}]$ ; the table drops a digit. With 0.0109 the real part of the Miki wavenumber at $f / σ = 0.01$ is $1.19$ instead of $2.88$ , inconsistent with the same table’s Delany-Bazley row ( $3.10$ from its own $r = 0.0862$ , $s = - 0.693$ ) and with the “modified Miki” row the book itself derives from it.
Evidence: digit check against the original Miki (1990) paper (Eqs. (30)–(34)) and cross-computation of both variants at the fit-range edge. Verified on PDF page 168 (printed p. 149) of Attenborough & Van Renterghem, Predicting Outdoor Sound 2e:2021, and on PDF page 4 (printed p. 22) of Miki, J. Acoust. Soc. Jpn (E) 11(1):1990.
Library behaviour: implements Miki’s original 0.109; the digitization point $f / σ = 0.1$ is pinned in tests/reference_data/ and in the conformance check “Miki 1990 Eqs. (30)-(34)”.
Status: unreported (textbook rather than a standard).

Attenborough & Van Renterghem, Predicting Outdoor Sound 2e (2021), Eq. (5.13)

Location: Eq. (5.13), the Johnson-Champoux-Allard bulk complex density, with $G (Λ) = 1 - 4 i T η ρ_{0} ω / (R_{S}^{2} Λ^{2} Ω^{2})$ .
The print: the tortuosity $T$ appears to the first power inside $G (Λ)$ .
The problem: Johnson et al. (1987) and the standard JCA formulation (Cox & D’Antonio 3e Eq. (6.19); Allard & Atalla) carry $T^{2} = α_{\infty}^{2}$ there. The first-power print breaks the high-frequency asymptote that defines the viscous characteristic length: with $T^{2}$ the density tends to $(T ρ_{0} /Ω) (1 + (1 - j) δ_{v} /Λ)$ with $δ_{v} = 2 η / ρ_{0} ω$ , while the printed form tends to a $δ_{v} / (Λ T)$ correction, which for $T = 2$ means an error of $29 %$ in the boundary-layer term for the same $Λ$ .
Evidence: asymptotic expansion of both variants against the Johnson et al. definition of $Λ$ and against Cox & D’Antonio Eq. (6.19); the library’s high-frequency JCA test pins the $T^{2}$ behaviour. Verified on PDF page 173 (printed p. 154) of Predicting Outdoor Sound 2e:2021.
Library behaviour: implements the standard $T^{2}$ form (Cox & D’Antonio Eq. (6.19)); the asymptote is pinned in test_high_frequency_density_asymptote.
Status: unreported (textbook rather than a standard).

Bies, Hansen & Howard, Engineering Noise Control 5e (2017), Eq. (8.141)

Location: Section 8.9.1, Eq. (8.141) (printed p. 461), the transmission loss of a muffler from the elements of its total four-pole matrix.
The print: $T L = 10 l g [(\frac{1 + M _{n}}{1 + M _{1}})^{2} \cdot \frac{1}{4} \cdot \frac{Z _{A 1}}{Z _{A n}} T_{11} + \frac{T _{12}}{Z _{A n}} + Z_{A 1} T_{21} + \frac{Z _{A n}}{Z _{A 1}} T_{22}^{2}],$ i.e. with the impedance ratio $Z_{A 1} / Z_{A n}$ weighting $T_{11}$ and its inverse weighting $T_{22}$ .
The problem: the source the equation itself cites (Munjal, Acoustics of Ducts and Mufflers 2e, Eq. (3.27), p. 105) carries the overall prefactor $Z_{A n} / Z_{A 1}$ (equivalently $S_{1} / S_{n}$ inside a $20 lo g_{10}$ form) with $T_{11}$ unweighted and $Z_{A 1} / Z_{A n}$ on $T_{22}$ . As printed, Eq. (8.141) fails the sudden-expansion limit: a zero-length element ( $T = I$ ) between $S_{1} = 0.01 m^{2}$ and $S_{n} = 0.02 m^{2}$ is a sudden area expansion with the classic $T L = 10 lo g_{10} [(1 + m)^{2} / (4 m)] = 0.512 dB$ ( $m = S_{n} / S_{1} = 2$ ), but the printed equation gives $\frac{1}{4} \cdot (Z_{A 1} / Z_{A n} + Z_{A n} / Z_{A 1})^{2} = 1.938 dB$ . Reading the ratios as an overall $Z_{A 1} / Z_{A n}$ prefactor instead is also wrong: it gives 6.532 dB on the same oracle and violates reciprocity ( $11.34$ vs -0.70 dB for an expansion chamber between unequal pipes; a negative TL for a passive element). The misprint is invisible whenever the inlet and outlet areas are equal, where every variant reduces to Eq. (8.148).
Evidence: numeric evaluation of the zero-length identity element and of an unequal-port expansion chamber under the printed form, the inverted prefactor and Munjal Eq. (3.27); only Munjal’s form reproduces the sudden-expansion classic (0.512 dB, both directions) and is reciprocal.
Library behaviour: transmission_loss in silencers.py implements Munjal Eq. (3.27), with the sudden-expansion limit and TL reciprocity pinned by regression tests (tests/noise_control/test_silencers.py) and a defensive note at the formula.
Status: unreported (textbook rather than a standard).

Long, Architectural Acoustics 2e (2014), Eq. (18.24) (sign of the microphone directivity)

Location: Chapter 18, “Multiple Open Microphones”, Eq. (18.24) (printed p. 699), the gain-before-feedback stability criterion generalised to several open microphones.
The print: $Z_{S} + L_{H - M} + Δ L_{nom} \leq L_{H - L} + D_{M} (θ) - 10$ , with the microphone directivity index entering the right-hand side with a plus sign.
The problem: Eq. (18.24) is the number-of-open-microphones generalisation of Eq. (18.20) (printed p. 698), which reads $Z_{S} + L_{H - M} \leq L_{H - L} - D_{M} (θ) - 10$ and which follows in turn from the oscillation condition Eq. (18.19), $Z_{S} + L_{H - M} = L_{H - L} - D_{M} (θ)$ , obtained by substituting the feedback-loop gain $G_{S} = L_{H - M} - L_{H - L} + D_{M} (θ)$ (Eq. (18.18)) into $Z_{S} + G_{S} = 0$ (Eq. (18.16)). Setting $N_{m} = 1$ makes $Δ L_{nom} = 0$ , so Eq. (18.24) must reduce to Eq. (18.20) and does not. The sign matters physically: $D_{M} (θ)$ is “usually negative” in Long’s own definition (about $- 2$ to -3 dB for a cardioid pointed at the talker), so as printed a directional microphone would cost gain before feedback instead of buying it, inverting the chapter’s own conclusion that “it is prudent to incorporate a cardioid or hypercardioid microphone into a system”.
Evidence: the printed equation reads $Z_{S} + L_{H - M} + Δ L_{nom} \leq L_{H - L} + D_{M} (θ) - 10$ , against $Z_{S} + L_{H - M} \leq L_{H - L} - D_{M} (θ) - 10$ two pages earlier, where the same position holds a minus. (An earlier revision of this entry quoted the pdftotext extraction, Z S þ L HM þ DL nom L HL þ D M ðqÞ 10, in which þ is the ligature this PDF uses for ”+” and every minus sign has been dropped entirely; that extraction cannot distinguish a plus from a minus and should never have been the evidence.) Verified on PDF page 697 (printed p. 699) and PDF page 696 (printed p. 698) of Long, Architectural Acoustics 2e (2014). The minus sign is the one that reproduces Long’s own worked special cases at $N_{m} = 1$ : with $Z_{S} = - 6 dB$ , Eq. (18.21) gives $L_{H - M} \leq L_{H - L} - D_{M} (θ) - 4$ (an omnidirectional microphone 4 dB below the average audience level), and Eq. (18.22) gives $L_{H - M} \leq L_{H - L} - 2$ for a cardioid at $D_{M} = - 2 dB$ . Neither special case is recoverable from the printed Eq. (18.24).
Library behaviour: feedback_stability in sound_reinforcement.py implements the sign of Eq. (18.20), with a note at the criterion. Both of Long’s special cases are pinned by regression tests (tests/electroacoustics/test_sound_reinforcement.py) and by the conformance checks “Long, Architectural Acoustics 2e, Eq. (18.21)” and “Eq. (18.22)”.
Status: unreported (textbook rather than a standard, so non-normative).

Long, Architectural Acoustics 2e (2014), Eq. (17.53) (constant of the communication bound)

Location: Chapter 17, “Restaurant Design”, Eq. (17.53) (printed p. 666), the minimum absorption per occupied table for adequate cross-table communication.
The print: $A_{tab} > 6.33 r_{s}^{2}$ .
The problem: the bound is Eq. (17.52), $L_{SN} = 10 lo g_{10} [Q / (4 π r^{2})] + 10 lo g_{10} [A_{tab} /4]$ , solved for $A_{tab}$ at the stated threshold $L_{SN} > - 6 dB$ , which gives $A_{tab} > 16 π \cdot 1 0^{- 0.6} r_{s}^{2} / Q$ . With the $Q = 2$ the chapter uses for a talker, that constant is 6.3130, not 6.33. The gap is $0.27 %$ , i.e. the last printed digit: 6.33 is what $16 π \cdot 1 0^{- 0.6} /2$ returns if $1 0^{- 0.6}$ is carried coarsely as 0.252 instead of 0.251 19. This is graded as a rounding-level discrepancy rather than a structural error of the formula, since the formula itself is confirmed by its companion (below) and no consistent alternative assumption reproduces 6.33 (it would require $Q = 1.995$ ).
Evidence: the immediately following Eq. (17.54) is the same closed form at the privacy threshold $L_{SN} < - 9 dB$ , and its printed constant 3.16 is exactly what $16 π \cdot 1 0^{- 0.9} /2 = 3.1640$ gives, confirming both the formula and $Q = 2$ . Only the -6 dB constant is off. What does not discriminate is Long’s prose one paragraph later, “at least 6.3 or more square meters (68 sq ft) of absorption per table”: 6.313 m² is $67.95 ft^{2}$ and 6.33 m² is $68.14 ft^{2}$ , so both print as 68 sq ft, and both round to 6.3 m². An earlier revision of this entry offered that conversion as corroboration. Verified on PDF page 665 (printed p. 666) of Long, Architectural Acoustics 2e (2014).
Library behaviour: absorption_per_table in crowd_noise.py computes the bound from Eq. (17.52) rather than hardcoding either constant, so both bounds stay mutually consistent; the 6.313 value and the printed 3.16 are pinned by regression tests (tests/room/test_crowd_noise.py) and the 3.16 constant by the conformance check “Long, Architectural Acoustics 2e, Eq. (17.54)”.
Status: unreported (textbook rather than a standard, so non-normative); graded as a rounding discrepancy rather than a structural defect.

Long, Architectural Acoustics 2e (2014), Table 14.7 (round elbow rows)

Location: Chapter 14, Table 14.7, “Insertion Loss of Round Elbows” (printed p. 541), indexed by the frequency-width product $f w$ (kHz times inches).
The print: four rows only: $f w < 1.9$ → 0 dB; $1.9 < f w < 3.8$ → 1 dB; $3.8 < f w < 7.5$ → 2 dB; $f w > 15$ → 3 dB.
The problem: the band $7.5 < f w < 15$ has no row at all, so the table jumps from $3.8 < f w < 7.5$ straight to $f w > 15$ . A duct-borne calculation lands in that band routinely: a $24 in$ elbow at 500 Hz has $f w = 12$ .
Evidence: the same data adapted from the same ASHRAE source appear in Bies, Hansen & Howard, Engineering Noise Control 5e, Table 8.11, indexed by $W / λ$ ( $= 0.074 f w$ ). Its round-elbow column has six rows, 0/1/2/3/3/3, and gives 3 dB for $0.55 \leq W / λ < 1.11$ , which is exactly the $7.5 < f w < 15$ band Long omits. Long’s four rows map onto Bies’ six as follows: the first three agree entry for entry, the fourth ( $f w > 15$ , 3 dB) legitimately merges Bies’ two identical top rows, and the band with no row is Bies’ fourth. An earlier revision of this entry said that “Tables 14.5 and 14.6 both carry six rows” and that “the other five rows of the two tables agree entry for entry”; on the page, Table 14.5 carries six rows and Table 14.6 five (it merges the same two identical top bands, legitimately), and Table 14.7 prints four, so neither count is right. Verified on PDF page 542 (printed p. 541) and PDF page 541 (printed p. 540) of Long, Architectural Acoustics 2e (2014).
Library behaviour: elbow_insertion_loss in hvac.py carries the six-row round column with 3 dB in the missing band, pinned by test_elbow_tables_by_frequency_width_product (tests/noise_control/test_hvac_long.py).
Status: unreported (textbook rather than a standard).

Long, Architectural Acoustics 2e (2014), Eq. 13.28 (units of U_G)

Location: Chapter 13, Eq. 13.28 (printed p. 521), the normalised pressure-drop coefficient $ξ = 334.9 \cdot Δ P / (ρ_{0} U_{G}^{2})$ of the diffuser sound-power model.
The print: the nomenclature under the equation gives “U_G = flow velocity prior to the diffuser (ft/min)” and, on the next line, ”= Q/(60·S_G) (for Q in cfm)”.
The problem: the two statements contradict each other. $Q$ in ft³/min divided by $60 S_{G}$ is a velocity in ft/s, not ft/min, and only the ft/s reading makes the constant right: $334.9/ ρ_{0}$ with $ρ_{0} = 0.075 lb/ft^{3}$ is $4465 \cdot Δ P / U^{2}$ , which is the standard velocity-pressure relation $Δ P / (U /4005)^{2}$ only when $U$ is converted from ft/s. Read as ft/min the coefficient comes out 3600 times too small. Eq. 13.27 itself declares $U_{G}$ in ft/s, so the “(ft/min)” label under Eq. 13.28 is the odd one out.
Evidence: dimensional check of $Q / (60 S_{G})$ ; reconstruction of the $334.9/ ρ_{0}$ constant from the velocity-pressure relation; and the peak frequency. What does not discriminate is the overall level: Eq. 13.27 carries $30 lo g_{10} ξ + 60 lo g_{10} U_{G}$ , and substituting Eq. 13.28 makes the velocity cancel identically, $30 lo g_{10} ξ + 60 lo g_{10} U_{G} = 30 lo g_{10} (334.9Δ P / ρ_{0})$ . For the Table 14.9 supply diffuser ( $S_{G} = 4 ft^{2}$ , $Q = 312$ cfm, $Δ P = 0.05$ in w.g.) both readings therefore return the same $L_{W} = 45.18 dB$ . An earlier revision of this entry claimed that the ft/min reading “misses it by 100 dB”, which is arithmetically impossible for a quantity that does not depend on the velocity at all. What does discriminate is Eq. 13.32, $f_{P} = 48.8 U_{G}$ , which is the only other place $U_{G}$ enters: read in ft/s the approach velocity is $1.3 ft/s$ and the peak falls at 63.4 Hz, i.e. in the 63 Hz octave, so the Eq. 13.31 shape puts 33.4 dB in that band against the printed 33; read in ft/min it is $78 ft/min$ , the peak moves to 3 806 Hz, and the same shape puts -8.2 dB in the 63 Hz band. Verified on PDF page 522 (printed p. 521) of Long, Architectural Acoustics 2e (2014).
Library behaviour: diffuser_sound_power in hvac.py reads $U_{G}$ in ft/s internally (SI at the interface), with the Table 14.9 row pinned by test_diffuser_sound_power_reproduces_the_table_14_9_row (tests/noise_control/test_hvac_long.py) and the conformance check “Long 2e Eqs. 13.27-13.33”.
Status: unreported (textbook rather than a standard).

Vigran, Building Acoustics (2008), Figure 8.37 caption (carpet stiffness exponent)

Non-normative source (textbook).
Location: section 8.4.2, the caption of Figure 8.37 on printed p. 320 / pdf p. 341, which labels the predicted improvement curves of two floor coverings laid on a heavyweight floor.
The print: “Predicted improvement with a linear model: stiffness of carpet squares 3.2·10^6 N/m, vinyl covering 5.2·10^6 N/m.” (Vigran writes the decimal separator as a period.)
The problem: the carpet exponent is one order too high. The body text introducing the figure, on printed p. 321, says of the carpet squares that “we have assumed that the covering has the same stiffness as used in Figure 8.36”, and Figure 8.36 is labelled $s = 3.2 \cdot 1 0^{5} N/m$ inside the plot, the same value the body text on printed p. 320 gives for it. The vinyl value in the same caption is correct.
Evidence: printed p. 320 states $3.2 \cdot 1 0^{5} N/m$ “giving a resonance frequency f0 of approximately 130 Hz with a hammer mass of 0.5 kg”, and $3.2 \cdot 1 0^{5} /0.5 / (2 π) = 127.3 Hz$ reproduces that while $3.2 \cdot 1 0^{6} /0.5 / (2 π) = 402.6 Hz$ is a frequency that appears nowhere in the section. The same arithmetic applied to the caption’s vinyl value gives $5.2 \cdot 1 0^{6} /0.5 / (2 π) = 513.3 Hz$ against the “approximately 510 Hz” printed on p. 321, which fixes the formula and the hammer mass the author used. Graphically, the two dashed prediction curves of Fig. 8.37 are about two octaves apart, matching the stiffness ratio $5.2 \cdot 1 0^{6} / (3.2 \cdot 1 0^{5}) = 16.25$ (a factor 4.03 in frequency) and not $5.2 \cdot 1 0^{6} / (3.2 \cdot 1 0^{6}) = 1.63$ (a factor 1.27). Verified on PDF page 341 (printed p. 320) of Vigran, Building Acoustics:2008, on which both caption exponents read 6 unambiguously and the body text of the same page reads $3.2 \cdot 1 0^{5} N/m$ , with the surrounding argument read on PDF page 340 (printed p. 319) and PDF page 342 (printed p. 321) of the same edition.
Library behaviour: none needed; the library takes the covering stiffness from the user through covering_contact_stiffness, and the printed cut-off frequencies it is anchored on come from Hopkins rather than from this caption.
Status: unreported.

Norton & Karczub, Fundamentals of Noise and Vibration Analysis for Engineers 2e (2003), Eq. (6.56)

Location: Section 6.6.1, Eq. (6.56), the coupling loss factor of two homogeneous plates joined by $N$ point connections (printed p. 418).
The print: the denominator bracket $(ρ_{s 1}^{2} h_{1}^{2} c_{L 1}^{2} + ρ_{s 2}^{2} h_{2}^{2} c_{L 2}^{2})$ appears to the first power.
The problem: as printed the expression is not dimensionless. The prefactor $4 N h_{1} c_{L 1} / (3 ω S_{1})$ already has the dimensions of $m^{2} s^{- 1}$ over $m^{2} s^{- 1}$ , i.e. unity, so the remaining ratio of the two bracketed products must be dimensionless too. That requires the sum to be squared, $A_{1} A_{2} / (A_{1} + A_{2})^{2}$ .
Evidence: the book’s own answer to problem 6.13 (printed p. 617). With the squared denominator the twelve-bolt aluminium pair gives $η_{12} = 1.43 \cdot 1 0^{- 2}$ at 125 Hz against the printed $1.44 \cdot 1 0^{- 2}$ , and matches the whole 125 Hz to 2 kHz column to better than $0.7 %$ ; with the printed (unsquared) denominator the result is not a loss factor at all. Verified on PDF page 438 (printed p. 418) of Norton & Karczub, Fundamentals of Noise and Vibration Analysis for Engineers 2e:2003.
Library behaviour: point_connection_coupling_loss_factor in junction_transmission.py implements the squared form, with the printed column pinned by a regression test (tests/vibration/structural/test_junction_transmission.py) and a note at the formula.
Status: unreported (textbook rather than a standard).

Norton & Karczub 2e (2003), problem 6.13 answer (eta_21 column)

Location: Answers to problems, problem 6.13 (printed p. 617), the two $η_{21}$ columns of the welded and bolted tables.
The print: for the two aluminium plates (plate 1: 3 mm, 2.5 m × 1.2 m; plate 2: 5.5 mm, 2.0 m × 1.2 m) the answer gives, at 125 Hz, $η_{21} = 5.77 \cdot 1 0^{- 3}$ (welded) and $2.64 \cdot 1 0^{- 2}$ (bolted).
The problem: both columns are exactly the corresponding $η_{12}$ column multiplied by $h_{2} / h_{1} = 1.833$ . The SEA consistency relationship is $n_{1} η_{12} = n_{2} η_{21}$ (Eq. 6.8) with the flat-plate modal density $n = S 12 / (2 c_{L} h)$ of Eq. (6.25), so the correct factor is $n_{1} / n_{2} = (S_{1} h_{2}) / (S_{2} h_{1}) = 2.292$ . The printed column drops the plate area ratio $S_{1} / S_{2} = 1.25$ .
Evidence: the ratio of the printed columns is 1.8333 to five digits in every band of both tables, which is $h_{2} / h_{1}$ exactly; the $η_{12}$ columns themselves reproduce from Eqs. (6.52) to (6.56) to better than 0.7 %. Verified on PDF page 637 (printed p. 617) of Norton & Karczub 2e:2003, the page that carries both answer tables.
Library behaviour: the $η_{12}$ columns are used as the regression oracle; $η_{21}$ is obtained from Eq. (6.8) with the full modal densities, and a test pins the 2.292 ratio explicitly (tests/vibration/structural/test_junction_transmission.py).
Status: unreported (textbook rather than a standard).

Norton & Karczub 2e (2003), problem 6.10 (platform area)

Location: Problems, problem 6.10 (printed pp. 593-594) and its answer (printed p. 617): a satellite platform coupled to an aluminium cylinder, 500 Hz octave, printed answers $η_{12} = 4.26 \cdot 1 0^{- 4}$ , $η_{21} = 3.92 \cdot 1 0^{- 4}$ and $Π_{in} = 1.31 W$ .
The print: the statement gives the aluminium platform as “5 mm thick and 3.5 m × 3 m”, i.e. 10.5 m².
The problem: that area is inconsistent with the three printed answers. Eq. (6.12) fixes $E_{1} / E_{2} = (η_{2} + η_{21}) / η_{12} = 6.554$ from the printed loss factors alone, whereas the stated geometry with the printed velocities (27.2 and 13.2 mm/s) gives 7.88. The energy ratio is independent of the modal densities and of the wave speed, so no choice of those can reconcile it; only the platform area can. The area the answers imply is 8.73 m², which is $3.5 \times 3$ minus the $π (0.75 m)^{2}$ footprint of the cylinder that Fig. P6.10 shows passing through the platform.
Evidence: with 8.73 m² the inversion of Eqs. (6.15), (6.8) and (6.10) returns $η_{12} = 4.256 \cdot 1 0^{- 4}$ , $η_{21} = 3.910 \cdot 1 0^{- 4}$ and $Π_{in} = 1.306 W$ , i.e. all three printed answers within 0.4 %; the cylinder’s own energy and modal density come out unchanged either way. Verified on PDF page 613 (printed p. 593), which carries the statement and its dimensions, and PDF page 637 (printed p. 617), which carries the three answers, of Norton & Karczub 2e:2003.
Library behaviour: power_injection_clf in experimental_sea.py implements the inversion as published; the regression test uses the free platform area and documents the discrepancy (tests/vibration/structural/test_experimental_sea.py).
Status: unreported (textbook rather than a standard).

Norton & Karczub 2e (2003), problem 3.14 (structural loss factor)

Location: Problems, problem 3.14 (printed p. 580) and its answer (printed p. 611): the octave-band transmission loss of a 20 mm particle board panel.
The print: the statement gives the panel a structural loss factor of “~1.5 × 10⁻²”; the answer gives 27 dB at 8 kHz and 38.6 dB at 16 kHz.
The problem: those two values are above the panel’s critical frequency (4885 Hz for Appendix 4 particle board, $f_{c} t = 97.7 m/s$ ) and therefore follow Cremer’s Eq. (3.110), which contains $10 lo g_{10} (η)$ . With $η = 1.5 \cdot 1 0^{- 2}$ the equation gives 37.0 dB and 48.5 dB, ten decibels above the printed answers; with $η = 1.5 \cdot 1 0^{- 3}$ it gives 27.0 dB and 38.5 dB.
Evidence: the 10 dB offset is exactly one decade of $10 lo g_{10} (η)$ , and the frequency dependence of the printed pair independently fixes $f_{c} = 4939 Hz$ against the Appendix 4 value of 4885 Hz. The eight values below coincidence reproduce exactly from Eq. (3.104) and do not involve $η$ . The discrepancy is a decade in a printed exponent, so the two figures were read as images rather than through the text layer. Verified on PDF page 600 (printed p. 580) and PDF page 631 (printed p. 611) of Norton & Karczub 2e (2003).
Library behaviour: the regression test uses $η = 1.5 \cdot 1 0^{- 3}$ , the value the printed answers require (tests/building/prediction/test_panel_transmission.py).
Status: unreported (textbook rather than a standard).

Vigran, Building Acoustics (2008), Eq. (9.18) (receiving-side coefficient)

Location: Section 9.2.3.2, Eq. (9.18) (printed p. 339), the transmission factor of the one-dimensional suspended-ceiling plenum model after Mechel (1980).
The print: the denominator reads $m_{S} L_{S} \cdot m_{R} L_{R} h$ with the unprimed $m_{R}$ , while the exponent of the same expression carries the primed $m_{R}^{'} = m_{R} + s_{R} τ_{R} / h$ of Eq. (9.17).
The problem: the two sides of the plenum are integrated the same way. The receiving-side integral is $\int_{0}^{L_{R}} exp (- ε m_{R}^{'} x) d x = (1 - exp (- ε m_{R}^{'} L_{R})) / (ε m_{R}^{'})$ , so the factor that normalises it must be $m_{R}^{'} L_{R}$ , exactly as the source-side one is $m_{S} L_{S}$ . Read literally, the printed expression is not a transmission factor at all: it carries a spurious $m_{R}^{'} / m_{R} = 1 + s_{R} τ_{R} / (h m_{R})$ , so it grows without bound as the plenum damping falls. Two consequences are visible with ordinary inputs ( $L_{S} = L_{R} = 5 m$ , $h = 0.6 m$ , $R_{S} = R_{R} = 25 dB$ , $ε = 2$ , $s_{S} = s_{R} = 0.5$ ): the model diverges as the plenum damping vanishes, giving $R_{cl} = 40.26 dB$ at $m_{R} = 0.01 1/m$ but only 26.48 dB at $1 0^{- 4}$ and 6.64 dB at $1 0^{- 6}$ , against the finite 40.85 dB that the derived reading returns for the same bare plenum, where the leakage term $s_{R} τ_{R} / h$ bounds the path; a plenum with no absorber at all is therefore predicted arbitrarily worse than the leak-limited value rather than equal to it. It also breaks energy conservation, returning $τ_{cl} = 4.45$ at $R_{S} = R_{R} = 6 dB$ , $m_{R} = 0.01$ and $τ_{cl} = 829$ at $R_{S} = R_{R} = 0 dB$ , $m_{R} = 1 0^{- 3}$ .
Evidence: with $m_{R}^{'}$ in the denominator every one of those pathologies disappears: $τ_{cl}$ flattens onto the leak-limited value as the damping vanishes, where the printed form keeps growing, and is bounded above by 1 because $(1 - exp (- ε m_{R}^{'} L_{R})) / (ε m_{R}^{'} L_{R}) \leq 1$ , and reduces to Vigran’s own small-attenuation result, Eq. (9.19) $τ_{cl} = ε^{2} τ_{S} τ_{R} L_{R} / (4 h)$ , whenever $m_{S} L_{S}$ and $m_{R}^{'} L_{R}$ are both small. With the printed $m_{R}$ the same limit picks up the factor $m_{R}^{'} / m_{R}$ , which diverges, so Eq. (9.18) as printed does not reduce to Eq. (9.19) at all: the two equations the book presents as a pair are inconsistent with each other. Verified on PDF page 361 (printed p. 339) of Vigran, Building Acoustics:2008, which shows the denominator carrying the unprimed $m_{R}$ while the exponent of the same expression carries the primed $m_{R}^{'}$ of Eq. (9.17).
Library behaviour: plenum_flanking_reduction_index in ceiling_plenum.py implements the derived $m_{R}^{'}$ in both the exponent and the denominator, with the reading documented at the formula, and rejects a transmission factor above unity rather than reporting a negative sound reduction index. Tests pin the physics the model owes (monotonicity in the damping, the $τ_{cl} \leq 1$ bound, the size of the Eq. (9.17) leakage term at a realistic ceiling) and the one property that separates the two readings: a bare plenum no worse than the undamped Eq. (9.20) value (tests/building/prediction/test_ceiling_plenum.py).
Status: unreported (textbook rather than a standard). Mechel’s original 1980 paper, which Vigran reproduces, was not available to check whether the misprint originates there.

Real Decreto 1367/2007, Annex IV A.3.3 (Kf and Ki threshold tables)

Location: Annex IV, section A.3.3, the $K_{f}$ (low-frequency) and $K_{i}$ (impulsive) correction tables, middle row of each.
The print: both tables print the 3 dB row as “Si 10 > Lf <= 15” and “Si 10 > Li <= 15” respectively (BOE-A-2007-18397, consolidated text).
The problem: the condition as printed is unsatisfiable. It reads “10 greater than Lf” and “Lf at most 15” simultaneously, which would select levels below 10 dB, but the row above it already assigns those to 0 dB (“Si Lf <= 10”) and the row below covers “Si Lf > 15”. The three rows only partition the range under the reading $10 < L_{f} \leq 15$ , so the ”>” is a typeset inversion of ”<”.
Evidence: the bracketing rows leave no other consistent reading; the identical construction appears in both tables, and the equivalent tables in the autonomous-community noise regulations that transpose this Annex print 10 < Lf <= 15. Verified on PDF page 26 (printed p. 26) of Real Decreto 1367/2007, BOE-A-2007-18397 consolidated text, on which the ”>” of both middle rows is unambiguous against the ”<=” glyphs of the same cell.
Library behaviour: low_frequency_correction and impulsive_correction implement $10 < L \leq 15$ , with a regression test pinning the three branches at the 10 dB and 15 dB boundaries.
Status: unreported (national regulation, not a standards body).

Commission Directive (EU) 2015/996, Annex II 2.2.1 (octave-band range of the road source)

Location: the Annex, point 2.2.1, second paragraph under the heading “Traffic flow” (OJ L 168, 1.7.2015, p. 8).
The print: “these sound power levels are calculated for each octave band i from 125 Hz to 4 kHz”.
The problem: the road source model contradicts its own coefficient database. Every band-dependent table of Appendix F, both in the 2015 text and in the version replaced by (EU) 2021/1226, is printed over the eight octave bands 63 Hz to 8 kHz (Table F-3 has no frequency columns at all), and point 2.1.1 of the same Annex defines the frequency range of the method as 63 Hz to 8 kHz. A calculation restricted to 125 Hz - 4 kHz would silently discard the 63 Hz and 8 kHz bands, which Appendix F tabulates like every other.
Evidence: corrected by the corrigendum published in OJ L 5, 10.1.2018, p. 35, which reads in full: ‘On page 8, in the Annex, in point 2.2.1, in the second paragraph under the heading “Traffic flow”: for: “each octave band i from 125 Hz to 4 kHz”, read: “each octave band i from 63 Hz to 8 kHz”’. The same corrigendum also adds “octave bands” to the frequency range of 2.1.1. Verified on PDF page 8 (printed p. L 168/8) of Commission Directive (EU) 2015/996:2015 for the printed restriction, on PDF page 1 (printed p. L 5/35) of the corrigendum for both items, and on PDF page 4 (printed p. L 168/4) and PDF page 124 (printed p. L 168/124) of the Directive for the conformant range.
Library behaviour: cnossos_road works over the corrected 63 Hz to 8 kHz grid (ROAD_OCTAVE_BANDS), pinned by test_octave_bands_are_the_corrected_range and by the workbook cases, whose published levels cover all eight bands.
Status: corrected by the issuing body (corrigendum of 10 January 2018); recorded because the uncorrected 2015 text is still the one most often downloaded and quoted.

Ainslie, Principles of Sonar Performance Modelling (2010), Eq. (9.57)

Textbook, not a standard.

Location: Section 9.1.1.2.4 (printed p. 457), the transition range between the mode-stripping and single-mode regimes of the Weston flux model.
The print: $r_{MS} \approx k^{2} H_{e}^{3} / (9 η)$ , where $H$ is the water depth, $H_{e}$ the Weston effective depth of Eq. (9.55), $k = ω / c_{w}$ and $η$ the reflection loss gradient.
The problem: the sentence immediately above it prescribes the derivation, “estimated by equating θ_n and θ_eff with n = 3/2”. The two angles are four printed pages apart, not on the same page as an earlier revision of this entry stated:
- Eq. (9.47), $θ_{eff} = (π H / (4 η r))^{1/2}$ , printed p. 453, with the true water depth $H$ (it comes from the multipath integral Eq. (9.46), whose $1/ (r H)$ prefactor is the cylinder area $A_{CS} = 2 π r H$ of Eq. (9.44), so $H$ is the depth that counts bottom bounces);
- Eq. (9.56), $θ_{n} \approx nπ / (k H_{e})$ , printed p. 457, with the effective depth $H_{e}$ (mode angles are set by the apparent pressure-release boundary).
Equating them at $n = 3/2$ gives $π H / (4 η r) = 9 π^{2} / (4 k^{2} H_{e}^{2})$ , that is $r_{MS} = k^{2} H_{e}^{2} H / (9 π η)$ . The printed form is larger by $π H_{e} / H$ . The factor $π$ is unconditional: it survives even if $H_{e}$ is substituted for $H$ in Eq. (9.47), which is presumably how the printed $H_{e}^{3}$ arose, and that reading would give $k^{2} H_{e}^{3} / (9 π η)$ , still $π$ below the print. The residual $H_{e} / H$ is the depth substitution itself, and it tends to 1 at high frequency. The other transition of the same section, Eq. (9.50) $r_{CS} = π H / (4 η ψ_{c}^{2})$ , follows its own derivation exactly (it is where Eq. (9.42) and Eq. (9.49) cross), so the defect is confined to Eq. (9.57).
Evidence: the symbolic re-derivation above, checked numerically for $H = 50 m$ , $f = 250 Hz$ , $c_{w} = 1500 m/s$ over the Table 9.1 sand seabed ( $η = 0.28 Np/rad$ , $ψ_{c} = 33.5 6^{\circ}$ , $H_{e} = 53.63 m$ , $k = 1.047 m^{- 1}$ ):

$r_{MS}$ $θ_{eff}$ there (Eq. 9.47)
derivation, $k^{2} H_{e}^{2} H / (9 π η)$ 19.9 km 4.808°
printed Eq. (9.57), $k^{2} H_{e}^{3} / (9 η)$ 67.1 km 2.619°

The ratio $67.1/19.9$ is $π H_{e} / H = 3.3695$ to every digit carried. The angle column is an independent check that does not depend on how the derivation is read: the first two mode angles of Eq. (9.56) are $θ_{1} = 3.20 5^{\circ}$ and $θ_{2} = 6.41 0^{\circ}$ , so $θ_{3/2} = 4.80 8^{\circ}$ . At the derived range the effective angle is exactly $θ_{3/2}$ , halfway between the first two modes, which is what the text asks for. At the printed range it has fallen to 2.619°, below $θ_{1}$ itself: the second mode would have been stripped long before, so that range cannot be where the single-mode regime begins. Both printed formulae are confirmed on PDF page 483 (printed p. 453) and PDF page 487 (printed p. 457) of Ainslie, Principles of Sonar Performance Modelling (2010).
Library behaviour: weston_regime_boundaries in propagation/weston_regimes.py implements the derivation-consistent $k^{2} H_{e}^{2} H / (9 π η)$ , which is also what keeps $θ_{eff}$ defined with $H$ everywhere the module evaluates Eq. (9.47). The equating rule is pinned by test_mode_stripping_boundary_equates_theta_eff_with_mode_3_over_2, which rebuilds both angles from the printed equations rather than from the implementation, and the shared definition of $θ_{eff}$ by test_composite_loss_and_the_boundary_use_the_same_effective_angle (both in tests/underwater/propagation/test_weston_regimes.py).
Status: unreported (textbook rather than a standard).

NMFS (2024) Updated Technical Guidance v3.0, Table 5 / Table ES2 (otariid C)

Regulatory guidance document, not a standard.

Location: Table 5 (printed p. 25), repeated as Table ES2 (printed p. 3) and again as Table 8 (printed p. 35): the auditory weighting parameter $C$ of the otariid pinniped in-water group (OW / OCW).
The print: $C = 1.37 dB$ .
The problem: the correct value is 1.36 dB. NMFS states so itself in the table’s own footnote: “During the public comment period, an error was identified with the Navy’s rounding, where this value should be 1.36, instead of 1.37. Because this is such a minor error and to remain consistent with the Navy, NMFS decided rely upon the value the Navy originally provided.” The document therefore knowingly publishes the wrong digit.
Evidence: independent recomputation of $C$ from its own definition, the negated peak of $W (f)$ , with the same row’s parameters $a = 1.58$ , $b = 5$ , $f_{1} = 2.53 kHz$ , $f_{2} = 43.8 kHz$ : $C = 1.3643 dB$ , which rounds to 1.36. The published weighted TTS onset of the same row ( $179 dB = K + C$ with $K = 178$ ) is unaffected by the third digit. The same recomputation reproduces every other row of the table to the printed two decimals, so the OW row is the only one that does not round from its own parameters. Verified on PDF page 36 (printed p. 25), PDF page 14 (printed p. 3) and PDF page 46 (printed p. 35) of NMFS Updated Technical Guidance v3.0:2024, all three carrying 1.37 with the identical footnote.
Library behaviour: bioacoustics/weighting.py implements 1.36 and keeps the printed 1.37 available as WeightingParameters.c_db_as_printed, so an assessment that must reproduce the published table verbatim still can. Pinned by test_nmfs_2024_otariid_c_uses_the_corrected_1_36.
Status: unreported (the issuing body has already documented it).

Southall et al. (2019), Aquatic Mammals 45(2), Table 7 (impulsive peak SPL)

Peer-reviewed journal paper, not a standard.

Location: Table 7 (printed p. 156), the impulsive-noise TTS and PTS onset criteria; the two in-air carnivore rows PCA and OCA.
The print: PCA TTS peak SPL 138 and PTS peak SPL 144; OCA TTS peak SPL 161 and PTS peak SPL 167 $dB re 20 μ Pa$ .
The problem: all four are typographical errors. The authors’ own errata (Aquatic Mammals 45(5), 569-572, DOI 10.1578/AM.45.5.2019.569) names all four on printed p. 569, “There are four typographical errors in Table 7 on page 156”, and reprints the corrected table on printed p. 570: PCA 155 and 161, OCA 170 and 176. The same errata also corrects the column headed “B” in Table 5 to the parameter b of Eq. (2), which it likewise calls a typographical error.
Evidence: the errata itself, which names each wrong value and its replacement, corroborated by the article’s own extrapolation rule. Note first what does not discriminate. The PTS peak = TTS peak + 6 dB rule of printed p. 155 is satisfied by the printed pair as well ( $144 - 138 = 6$ , just as $161 - 155 = 6$ ), so it says nothing about which pair is right. Nor does the duplication visible in the printed rows, where the peak-SPL TTS entry equals that same row’s PTS-onset SEL entry (PCA 123 / 138 / 138 / 144 and OCA 146 / 161 / 161 / 167, reading TTS SEL, TTS peak, PTS SEL, PTS peak): for these two in-air rows that equality is forced by two rules the article states on printed p. 155, both adding 15 dB to the same base TTS SEL, so it would hold whatever the SEL values were. An earlier revision of this entry read that equality as the signature of a column slip; it is instead the printed table being internally consistent with the article’s own in-air method, which is what makes the errata the only thing that settles the matter.
- Value. The corrected numbers are close to what the article’s extrapolation rule produces, with the caveat that the rule is not stated for these rows. Printed p. 155 sets the impulsive peak-SPL TTS onset of a group without direct data at the hearing threshold at the frequency of best sensitivity $f_{0}$ plus 159 dB, and restricts that rule explicitly to the in-water groups: “For other species groups in water (LF, SI, PCW, and OCW), 159 dB was added to the value of the hearing threshold at f₀”. It works the rule through for PCW: “Peak SPL TTS onset was estimated as 212 dB re 1 µPa (53 dB at f₀ + 159 dB)”. Evaluating the Table 2 group audiogram at the Table 4 $f_{0}$ reproduces the three in-water rows the errata does not touch (SI 219.6 against a published 220; PCW 212.5 against 212; OCW 226.1 against 226), which validates the rule where the article applies it. Extending it to the two in-air carnivore rows, which the article does not do, gives PCA $- 4.6 dB re 20 μ Pa$ at 2.3 kHz and OCA $11.4 dB re 20 μ Pa$ at 10 kHz, hence 154.4 and 170.4. Those reproduce the corrected 155 and 170 to within 0.6 dB and are 16 dB and 9 dB away from the printed 138 and 161, which is what makes them corroborating rather than confirming; note that 154.4 rounds to 154, not to 155, and an earlier revision of this entry claimed that it rounded to the corrected value.
- A second, unrepaired inconsistency. Printed p. 155 states that for the in-air carnivores specifically “a nominal 15 dB offset is used … between the SEL-based TTS threshold and the peak SPL-based threshold”, which reproduces the printed 138 and 161 from the SEL column. That sentence, not the +159 dB rule, is the one the article’s own method applies to PCA and OCA. The errata resolves the conflict in favour of values consistent with the +159 dB rule, so it supersedes the sentence as well as the table; the sentence is left standing in the article. Verified on PDF page 31 (printed p. 155) and PDF page 32 (printed p. 156) of Southall et al. (2019), Aquatic Mammals 45(2), which carry the “in water (LF, SI, PCW, and OCW)” restriction, the 15 dB in-air offset in the same paragraph, both statements of the +6 dB rule, and the article’s Table 7 with the PCA row 123 / 138 / 138 / 144 and the OCA row 146 / 161 / 161 / 167. The errata is a publication of its own, Aquatic Mammals 45(5), 569-572, bound at the end of the copy the authors distribute: verified there on PDF page 109 (printed p. 569), which names all four values and their replacements, and PDF page 110 (printed p. 570), which reprints Table 7 with PCA 123 / 155 / 138 / 161 and OCA 146 / 170 / 161 / 176.
Library behaviour: the errata-corrected values are the ones implemented in bioacoustics/weighting.py, pinned by test_southall_table_7_errata_values_are_implemented, with the +159 dB rule itself checked against the audiogram in test_southall_impulsive_peak_spl_is_threshold_at_f0_plus_159_db for the in-water groups the article restricts it to and, separately and with the extrapolation labelled as such, for PCA and OCA.
Status: reported by the authors themselves (errata published 2019).

Directive (EU) 2015/996, Annex II 2.3.2 (roughness conversion in km/h)

Location: the “Definition” paragraph of Wheel and rail roughness (OJ L 168, 1.7.2015, p. 19) and the first paragraph after formula (2.3.11) (p. 21).
The print: “it shall be converted to a frequency spectrum f = v/λ, where f is the centre band frequency of a given 1/3 octave band in Hz, λ is the wavelength in m, and v is the train speed in km/h”, and, for impact noise, “using the relation λ = v/f, where f is the 1/3 octave band centre frequency in Hz and v is the s-th vehicle speed of the t-th vehicle type in km/h”.
The problem: dimensionally impossible. A frequency in hertz is a speed in metres per second divided by a wavelength in metres; reading the speed in km/h into $f = v / λ$ multiplies every frequency by 3,6, placing the whole roughness spectrum a factor 3,6 too high in frequency, which is more than an octave and a half.
Evidence: verified on PDF page 19 (printed p. L 168/19) and PDF page 21 (printed p. L 168/21) of Directive (EU) 2015/996. The corrigendum of OJ L 5, 10.1.2018, p. 35 replaces “km/h” by “m/s” in both places. The flow equation (2.3.2) genuinely does take its speed in km/h, which is what makes the misprint plausible.
Library behaviour: roughness_to_frequency converts the speed to m/s before dividing, as corrected, and its docstring says so. The reference implementation the Commission published with the source module does the same, and the 123 committed workbook cases would not reproduce otherwise.
Status: unreported (corrected by the issuing body in 2018).

Directive (EU) 2015/996, Appendix G, Table G-1, second table (wrong symbol)

Location: Table G-1, “Coefficients Lr,TR,i and Lr,VEH,i for rail and wheel roughness”, second table (OJ L 168, 1.7.2015, pp. 130-131).
The print: the second table is headed $L_{r, V E H, i}$ , the same symbol as the first.
The problem: its two columns are “EN ISO 3095:2013 (Well maintained and very smooth)” and “Average network (Normally maintained smooth)”, which are the rail-roughness classes E and M of digit 2 of the track descriptor in Table [2.3.b]. The table is the rail roughness $L_{r, T R, i}$ , the quantity the table’s own title announces and which is otherwise missing from Appendix G.
Evidence: verified on PDF page 130 (printed p. L 168/130) of Directive (EU) 2015/996:2015, the page carrying the header of the second table. The corrigendum of OJ L 5, 10.1.2018 re-titles it $L_{r, T R, i}$ , and Commission Delegated Directive (EU) 2021/1226 Annex point (20)(a) reprints it under that symbol when it replaces it, verified on PDF page 35 (printed p. L 269/99) of that Directive.
Library behaviour: rail_roughness returns the second table of G-1 as the rail roughness of (2.3.7) and wheel_roughness returns the first as the wheel roughness, which is the only assignment under which the classes of Table [2.3.b] can be reached at all.
Status: unreported (corrected by the issuing body in 2018).

Directive (EU) 2015/996, Appendix G, Table G-5, 6 350 Hz row (50 dB notch)

Location: Table G-5, “Coefficients LW,0,idling for traction noise”, the 6 350 Hz row of the “Diesel locomotive (c. 2 200 kW)” pair (OJ L 168, 1.7.2015, p. 138).
The print: Source A 31,4 dB and Source B 30,7 dB.
The problem: both are about 50 dB below their own neighbours in the same column: 90,5 / 89,5 dB at 5 000 Hz and 81,2 / 80,6 dB at 8 000 Hz. No physical traction source has a 50 dB notch one third of an octave wide, and no other column of the table has anything comparable. The leading digit 8 was lost.
Evidence: verified on PDF page 138 (printed p. L 168/138) of Directive (EU) 2015/996:2015, which carries the 5 000, 6 350 and 8 000 Hz rows and the “Diesel locomotive (c. 2 200 kW)” column header. Commission Delegated Directive (EU) 2021/1226 Annex point (20)(f), verified on PDF page 39 (printed p. L 269/103) of that Directive, replaces the 4th column, 25th row by “81,4” and the 5th column, 25th row by “80,7”, restoring the monotone roll-off. The same two values appear as 31,41 and 30,71 in the IMAGINE catalogue file the Commission distributes with its reference source module, so the error predates the Directive.
Library behaviour: ships the corrected 81,4 / 80,7 and pins them, together with the assertion that neither value is more than 10 dB from either neighbour, in test_table_g5_carries_the_2021_correction_at_6300_hz.
Status: unreported (corrected by the issuing body in 2021).

Directive (EU) 2015/996, Appendix G, band and wavelength labels

Location: the frequency column of Tables G-3, G-5 and G-6 and the wavelength column of Table G-1 (OJ L 168, 1.7.2015, pp. 129-140).
The print: the 1/3-octave band centres are labelled 316 Hz, 3 160 Hz and 6 350 Hz, and the wavelengths 120 mm, 12 mm, 3,2 mm and 1,2 mm.
The problem: neither series is the preferred one. The nominal 1/3-octave centres of IEC 61260-1 are 315, 3 150 and 6 300 Hz, and the R10 preferred numbers around those wavelengths are 125, 12,5, 3,15 and 1,25 mm. The Commission’s own catalogue files, distributed with the reference source module, use the preferred wavelength series throughout.
Evidence: verified on PDF pages 129, 130, 131, 133, 134, 135, 137 and 138 (printed pp. L 168/129 to L 168/138) of Directive (EU) 2015/996:2015, which carry every occurrence of the four wavelengths and of the three band labels. Commission Delegated Directive (EU) 2021/1226 Annex point (20)(c) replaces the $L_{H, T R, i}$ section of Table G-3 outright and points (20)(d), (f) and (g) replace the three frequency labels in the remaining sections and in Tables G-5 and G-6; the tables it replaces outright carry the preferred wavelengths. But point (20)(a) replaces only “the second table” of Table G-1, so the wavelength labels 120, 12, 3,2 and 1,2 mm still stand on the first table of G-1, the wheel roughness $L_{r, V E H, i}$ , which is the one table that keeps them.
Library behaviour: the frequency grid is the IEC 61260-1 one throughout. The wavelength grids are kept as printed, one per table, and each roughness spectrum is resampled on its own grid rather than forced onto a common one, which is what _WAVELENGTHS_WHEEL and _WAVELENGTHS_STANDARD are for; the difference between the two is pinned by test_wheel_roughness_keeps_the_non_standard_wavelength_grid.
Status: unreported (frequency labels corrected by the issuing body in 2021; the wheel-roughness wavelength labels stand).

Directive (EU) 2015/996, Annex II 2.3.2, curve squeal (unassigned endpoints)

Location: the Squeal paragraph (OJ L 168, 1.7.2015, p. 21).
The print: “The emission level to be used is determined for curves with radius below or equal to 500 m and for sharper curves and branch-outs of points with radii below 300 m”, and then “squeal noise shall be considered by adding 8 dB for R < 300 m and 5 dB for 300 m < R < 500 m”.
The problem: the two open intervals leave $R = 300 m$ and $R = 500 m$ with no excess at all, and $R = 500 m$ is explicitly inside the scope the same paragraph has just set. A 500 m curve therefore falls out of a rule written to include it.
Evidence: verified on PDF page 21 (printed p. L 168/21) of Directive (EU) 2015/996:2015, on which both inequalities of the rule sentence are strict while the scope sentence above them reads “below or equal to 500 m”. Commission Delegated Directive (EU) 2021/1226 Annex point (4)(b), verified on PDF page 4 (printed p. L 269/68) of that Directive, replaces the paragraph with a table whose intervals are closed, “R <= 300 m” and “300 m < R <= 500 m”.
Library behaviour: curve_squeal_excess implements the 2021 table, so $R = 300 m$ returns 8 dB and $R = 500 m$ returns 5 dB; the boundaries are pinned in test_curve_squeal_rule_of_2021.
Status: unreported (corrected by the issuing body in 2021).

Allard & Atalla, Propagation of Sound in Porous Media 2e (2009), Eq. (6.85)

Textbook, not a standard.

Location: Sect. 6.5.2 (printed p. 123), the second form of the shear-wave velocity ratio $μ_{3}$ .
The print: $μ_{3} = (N δ_{3}^{2} - ω^{2} ρ_{11}) / (ω^{2} ρ_{22})$ , offered as an alternative to Eq. (6.84), $μ_{3} = - ρ_{12} / ρ_{22}$ .
The problem: the two printed forms are not equal. Substituting the shear wavenumber of Eq. (6.83), $δ_{3}^{2} = (ω^{2} / N) (ρ_{11} ρ_{22} - ρ_{12}^{2}) / ρ_{22}$ , into the printed Eq. (6.85) gives $- ρ_{12}^{2} / ρ_{22}^{2}$ , which is Eq. (6.84) multiplied by the spurious factor $ρ_{12} / ρ_{22}$ . The denominator should read $ω^{2} ρ_{12}$ .
Evidence: the book’s own derivation. Eq. (6.80), printed p. 122, is $- ω^{2} ρ_{11} ψ_{s} - ω^{2} ρ_{12} ψ_{f} = N \nabla^{2} ψ_{s} = - N δ_{3}^{2} ψ_{s}$ , so $(N δ_{3}^{2} - ω^{2} ρ_{11}) ψ_{s} = ω^{2} ρ_{12} ψ_{f}$ and therefore $μ_{3} = ψ_{f} / ψ_{s} = (N δ_{3}^{2} - ω^{2} ρ_{11}) / (ω^{2} ρ_{12})$ . With that reading the two forms agree identically wherever $ρ_{12}$ is non-zero; at $ρ_{12} = 0$ the corrected quotient is $0/0$ while Eq. (6.84) stays defined and gives $μ_{3} = - ρ_{12} / ρ_{22} = 0$ , which is the value to use there. The printed form instead differs from Eq. (6.84) by the factor $ρ_{12} / ρ_{22}$ , so it coincides with it only where that ratio is exactly 0 or exactly 1. With $ρ_{12} / ρ_{22} = ρ_{0} / (ϕ ρ_{eq}) - 1$ those two cases ask for $ρ_{eq} = ρ_{0} / ϕ$ and $ρ_{eq} = ρ_{0} / (2 ϕ)$ , both real; the effective density of a lossy porous medium is complex, so neither is ever met. Verified on PDF page 132 (printed p. 123) of Allard & Atalla, Propagation of Sound in Porous Media 2e:2009, which carries both printed forms, and on the facing page for Eq. (6.80).
Library behaviour: biot_waves implements Eq. (6.84) as printed, and test_shear_velocity_ratio_matches_the_corrected_second_printed_form checks it against the corrected Eq. (6.85) over four decades of frequency, and also asserts that the form exactly as printed disagrees.
Status: unreported.

Allard & Atalla 2e (2009), Eq. (11.48) and Table 11.1 (poroelastic layer)

Textbook, not a standard.

Location: Sect. 11.3.3 (printed pp. 251-252), the fluid normal stress $σ_{33}^{f}$ of a poroelastic layer and the matrix $[Γ]$ it feeds.
The print: Eq. (11.48) reads
$σ_{33}^{f} = i \sum (Q + R μ_{i}) (k_{t}^{2} + k_{i 3}^{2}) {- (A_{i} - A_{i}^{'}) cos (k_{i 3} x_{3}) + j (A_{i} - A_{i}^{'}) sin (k_{33} x_{3})}$
and Table 11.1 writes $k_{i 3}$ in the two columns that carry $μ_{1}$ , $D_{1}$ and $E_{1}$ .
The problem: two independent misprints in the same equation, plus a subscript slip in the table.
- The coefficient of the symmetric amplitude $(A_{i} + A_{i}^{'})$ is missing: Eq. (11.48) attaches both terms to $(A_{i} - A_{i}^{'})$ , which would leave the first and third columns of $[Γ]$ with no $σ_{33}^{f}$ entry at all, contradicting Table 11.1, whose row 6 prints $- E_{1} cos (k_{13} x_{3})$ and $- E_{2} cos (k_{23} x_{3})$ in exactly those columns. The first term is $- (A_{i} + A_{i}^{'}) cos (k_{i 3} x_{3})$ .
- The sine carries $k_{33}$ , the shear wave-number component, inside a sum over the two compressional waves $i = 1, 2$ . It must be $k_{i 3}$ . Table 11.1 again gives the intended reading: its row 6 has $j E_{1} sin (k_{13} x_{3})$ and $j E_{2} sin (k_{23} x_{3})$ , and zero in both shear columns, because a shear wave produces no dilatation and therefore no $σ_{33}^{f}$ .
- Table 11.1 prints the running subscript $k_{i 3}$ in its first two columns, which belong to the first compressional wave alone: the $μ_{1}$ , $D_{1}$ and $E_{1}$ in the same columns make $k_{13}$ the only consistent reading.
Evidence: the two readings above are forced by Table 11.1, which the same page declares to be the tabulation of Eqs. (11.37), (11.38) and (11.46)-(11.48). They are also what the stress-strain relation Eq. (11.41), $σ_{33}^{f} = R div u_{f} + Q div u_{s}$ , gives when the displacement potentials of Eqs. (11.22)-(11.25) are differentiated directly. Verified on PDF page 257 (printed p. 251) and PDF page 258 (printed p. 252) of Allard & Atalla, Propagation of Sound in Porous Media 2e:2009, which carry Eq. (11.48) and the two Table 11.1 columns as printed.
Library behaviour: the $[Γ]$ of Table 11.1 is implemented with the corrected readings, and test_gamma_matches_the_field_rebuilt_from_the_potentials checks all thirty-six of its entries at three frequencies, three depths and three angles of incidence against the field rebuilt from Eqs. (11.22)-(11.28) without going through the table.
Status: unreported.

Allard & Atalla 2e (2009), Sect. 6.6.3 (thickness of the second sample)

Textbook, not a standard.

Location: Sect. 6.6.3, printed p. 129, the two glass-wool samples whose measured and predicted surface impedances are Figures 6.10 and 6.11.
The print: the first sentence says the impedances are shown “for l = 10 cm and l = 5.4 cm”; two sentences later the peak of the second sample is placed at “860 Hz for l = 5.6 cm”, and the caption of Figure 6.11 says “l = 5.6 cm”.
The problem: the two thicknesses cannot both be right.
Evidence: textual, and only textual. Two printed statements carry 5.6 cm, the sentence about the 860 Hz peak and the independent caption of Figure 6.11, against one carrying 5.4 cm; a single slip in the opening sentence is the shorter explanation than the same slip made twice. The numbers do not settle it, and this entry does not claim they do. The book gives no peak-finding rule, and the answer follows the rule chosen:
- Taking the peak as the maximum of $Im (Z_{s})$ , Eq. (6.107) on the fully specified Table 6.1 glass wool gives 863.5 Hz for 5.6 cm (+0.4 % against the printed 860) and 896.2 Hz for 5.4 cm (+4.2 %), which favours 5.6 cm. But the same rule puts the undisputed 10 cm sample at 480.0 Hz against its printed 470, a +2.1 % bias of the same size as the effect being resolved.
- Taking the peak as the maximum of $∣ Z_{s} - Z_{s, rigid} ∣$ , which is the departure the same paragraph describes (“close to each other, except around the peaks which are not predicted by the one-wave model”), the 10 cm sample lands at 469.2 Hz (-0.2 %) and both printed frequencies then come out of the pair (10 cm, 5.4 cm): 861.2 Hz for 5.4 cm (+0.1 %) against 831.0 Hz for 5.6 cm (-3.4 %). That rule favours 5.4 cm.
- Scaling the 10 cm peak is no help either, and leans the other way from the conclusion: $470 \times (10/5.4) = 870 Hz$ is 10 Hz from the published 860, $470 \times (10/5.6) = 839 Hz$ is 21 Hz from it.
- The agreement of “860 Hz” with “5.6 cm” is in any case partly circular, since both sit in the same clause: it tests that sentence against itself, not which of the two sentences is the misprint. Verified on PDF page 138 (printed p. 129) of Allard & Atalla, Propagation of Sound in Porous Media 2e:2009, on which the lone 5.4 cm and the 5.6 cm of the 860 Hz clause sit on the same page, and on PDF page 139 (printed p. 130) of the same edition for the Figure 6.11 caption, the second sentence carrying 5.6 cm.
Library behaviour: recorded, with no effect on the implementation. test_impedance_peak_of_the_thin_layer_resolves_the_printed_thickness pins the 5.6 cm peak against the published 860 Hz under the $Im (Z_{s})$ rule and checks that the 5.4 cm reading is the worse of the two under that rule.
Status: unreported, and the weakest of the four entries here: the conclusion rests on the two-against-one reading of the printed page, not on a computation.

Allard & Atalla 2e (2009), Sect. 6.5.4 (the frame-borne velocity ratio)

Textbook, not a standard.

Location: Sect. 6.5.4, printed p. 125, the one sentence of the book that quotes computed values of $μ_{b}$ for the Table 6.1 glass wool.
The print: “The ratio modulus $∣ μ_{b} ∣$ of the velocities of the frame and the air for the frame-borne wave decreases from 1.0 at 50 Hz to 0.82 at 1500 Hz.”
The problem: the two quoted values are the real part of $μ_{b}$ , not its modulus. $μ_{b}$ is complex, and the sentence names the modulus explicitly.
Evidence: on the fully specified Table 6.1 material the model gives $μ_{b} (1500 Hz) = 0.811 + 0.473 j$ . Its real part is 0.811, 1.1 % from the printed 0.82; its modulus is 0.939, 14.5 % away. Read as the real part, the sentence is right at both ends and describes a monotone decrease: $Re (μ_{b})$ is 1.002 at 50 Hz and passes through 0.82 at 1467 Hz, 2.2 % from the printed 1500 Hz. Read as the modulus it is right at neither: $∣ μ_{b} ∣$ is 1.002 at 50 Hz but rises to 1.008 by 400 Hz before turning over, and only reaches 0.82 at 2634 Hz, 76 % above the printed frequency. No admissible reading of the printed inputs closes that gap. With the loss factor at 0 or at 0.2, the viscous length halved or doubled, $Λ^{'} = 2Λ$ in place of the printed $1.1 \cdot 1 0^{- 4} m$ , the resistivity halved or doubled, the tortuosity at 1 or the Poisson coefficient at 0.3, $∣ μ_{b} (1500) ∣$ moves only between 0.874 and 1.073. The closest of the eight, 0.874 at zero loss factor, is still 6.6 % from the printed 0.82, and it loses the 495 Hz branch crossing of the same section altogether; the only variant that keeps that crossing ( $Λ^{'} = 2Λ$ , 495.2 Hz) leaves $∣ μ_{b} ∣$ at 0.937. Reading the sentence as $Re (μ_{b})$ needs no variant at all. Verified on PDF page 134 (printed p. 125) of Allard & Atalla, Propagation of Sound in Porous Media 2e:2009, which carries the sentence and its 0.82.
Library behaviour: biot_waves computes $μ_{b}$ from Eq. (6.71) as printed. The conformance row and test_frame_borne_velocity_ratio_matches_the_two_published_values are written against $Re (μ_{b})$ , and say so.
Status: unreported.

Recorded here to prevent future “fixes” that would break agreement with the published sources:

ISO 12354-1:2017 Table L.8 / ISO 12354-2:2017 Table G.8, first row: the row labelled “Int. wall 1/2 – Ext. wall 1/2” prints $m_{i}^{'} = 219, 0 kg/m^{2}$ and $m_{⊥ i}^{'}$ (Part 2: $m_{orthogonal}^{'}$ ) $= 360, 0 kg/m^{2}$ , which is the assignment for a path leaving the external wall, the opposite of the direction the row’s own label gives. Read in the row’s direction the element carrying the path is the internal wall, so $m_{i}^{'}$ should be 360,0 and the perpendicular mass 219,0. It is a labelling slip and nothing else: the branch is the rigid-T corner branch $K_{12} = 5, 7 + 5, 7 M^{2}$ , where only $M^{2}$ enters, so both assignments return the same 5,965 → 6,0 dB. The second row of each table, “Ext. wall 1/2 – Ext. wall 1/2”, is the through branch $5, 7 + 14, 1 M + 5, 7 M^{2}$ , where the sign of $M$ does matter, and it is labelled and populated consistently ( $M = lo g_{10} (360/219)$ gives 9,006 → the printed 9,0). Verified on PDF page 89 (printed p. 83) of ISO 12354-1:2017 and PDF page 46 (printed p. 40) of ISO 12354-2:2017. Not registered as an erratum because no number depends on it; registered here so that a future reader does not “correct” the library’s per-path convention to match the printed row.
Francois-Garrison pure-water term: the two published $A_{3}$ cubics do not meet exactly at the 20 °C switch (a step of $1 \cdot 1 0^{- 7} f^{2} dB/km$ , 0.1 dB/km at 1 MHz). Inherent in the published coefficients.
Ainslie-McColm simplification: the paper’s “within 10 % of Francois-Garrison” claim is marginally exceeded at the extreme corners of its stated domain (10.4 % at −6 °C / 1 MHz; 12.3 % at 7 km depth). A property of the published fit; both transcriptions verified digit-for-digit.
CNOSSOS-EU Annex II 2.3, missing equation number: the railway section numbers its formulae (2.3.1), (2.3.2), (2.3.4), (2.3.5)…, with no (2.3.3) anywhere in Annex II. Verified on PDF page 17 (printed p. L 168/17) of Directive (EU) 2015/996:2015, where (2.3.2) and (2.3.4) sit one above the other. Nothing is missing from the method; only the numbering skips.
CNOSSOS-EU corrigendum of 2018, Table G-3 column codes: the corrigendum is reported to head the seven $L_{r, T R}$ columns “B/S B/M B/H B/S B/M B/H B/H”, where the first three should read “M/S M/M M/H” and the last “W”, and Commission Delegated Directive (EU) 2021/1226 Annex point (20)(c) does replace that header with the corrected codes plus a new column D. It is left unregistered because the corrigendum itself is published only as HTML on EUR-Lex, so no printed page of it could be obtained here, and this registry does not record a claim about a printed symbol that has not been read off the page. The 2015 print of the same table, which was read, carries descriptive headers (“Mono-block sleeper on soft rail pad” and so on) and no defect.
Long, Architectural Acoustics 2e, Chapter 17, adjacent-table level: the restaurant example states that “at an adjacent table 3 m (10 ft) away, the direct field level from our conversation is about 54 dB”, where his own Eq. (17.50) with the $Q = 2$ and $L_{W} = 70 dB$ that yield his 60 dB at 1.2 m gives 52.5 dB. It is left unregistered because the intended reading cannot be established from the book: 54 dB is also what the same equation gives at 2.5 m (54.1 dB, and 2.5 m is the table spacing the next paragraph derives), and what a single 6 dB distance doubling from the rounded 60 dB would give, while the printed “3 m (10 ft)” is self-consistent in both units and is repeated in the preceding paragraph. speech_direct_level evaluates Eq. (17.50) as printed, so it returns 52.5 dB there; do not “correct” it toward 54 dB.
ICAO Annex 16 EPNL constant: the Annex’s rounded constant 13 for uniform 0.5 s records differs from the exact $- 10 lo g_{10} (T_{0})$ form by 0.0103 dB; the library uses the exact form, which the ETM’s integrated reference reproduces to five decimals.
Long Table 14.9 element rows: the worked duct-borne sheet of Chapter 14 was produced by a commercial program, as the text introducing it states, and several of its element rows do not follow from the tables printed beside them: the fan row (90/86/82/79/77/75/71/61 dB) is not what Eq. 13.1 gives with the Table 13.5 forward-curved constants at that duty (99/99/89/84/82/77/72/67 dB, and not a level shift of it), and the flexible-duct row (14/14/16/15/17/22/16/13 dB) is not the Table 14.4 entry for 12 in by 6 ft (3/5/10/15/17/16/9 dB). The library implements the printed equations and tables, and uses the sheet only for what it genuinely pins, the cascade arithmetic; its element rows are fed in as published in tests/noise_control/test_duct_path.py. The sheet’s own rounding is likewise not always self-consistent (supply row 3 prints a Sum of 49 dB at 500 Hz where $76 - 28 = 48$ , then a Combined consistent with 48), which is why the comparison runs at the 1 dB the printed sheet carries.

Creado y mantenido por José Manuel Requena Plens ORCID Todos los proyectos

	$r_{MS}$	$θ_{eff}$ there (Eq. 9.47)
derivation, $k^{2} H_{e}^{2} H / (9 π η)$	19.9 km	4.808°
printed Eq. (9.57), $k^{2} H_{e}^{3} / (9 η)$	67.1 km	2.619°